Study of Homeosis in the Flower of Philodendron (Araceae): a Qualitative and Quantitative Approach

Abstract

This study deals specifically with floral organogenesis and the development of the inflorescence of Philodendron squamiferum and P. pedatum. Pistillate flowers are initiated on the lower portion of the inflorescence and staminate flowers are initiated on the distal portion. An intermediate zone consisting of sterile male flowers and atypical bisexual flowers with fused or free carpels and staminodes is also present. This zone is located between the sterile male and female floral zones. In general, the portion of bisexual flowers facing the male zone forms staminodes, and the portion facing the female zone develops an incomplete gynoecium with few carpels. The incomplete separation of some staminodes from the gynoecial portion of the whorl shows that they belong to the same whorl as the carpels. There are two levels of aberrant floral structures in Philodendron. The first one is represented by the presence of atypical bisexual flowers, which are intermediates between typical female flowers and typical sterile male flowers. The second one is the presence of intermediate structures between typical carpels and typical staminodes on a single atypical bisexual flower. The atypical bisexual flowers of P. squamiferum and P. pedatum are believed to be a case of homeosis where carpels have been replaced by sterile stamens on the same whorl. A quantitative analysis indicates that in both species, on average, one staminode replaces one carpel.

INTRODUCTION

The phenomenon of homeosis, defined in plants as the complete or partial replacement of one organ by a different organ (Lehman and Sattler, 1992), is believed to play an important role in the ontogeny and phylogeny of reproductive organs. This topic has recently been discussed extensively by morphologists (Sattler, 1988, 1994; Mabberley and Hay, 1994; Endress, 1996) and molecular biologists (Frohlich and Meyerowitz, 1997; Kramer et al., 1998; Mouradov et al., 1998; Purugganan, 1998; Winter et al., 1999; Becker et al., 2000; Kramer and Irish, 2000). Although the concept of homeosis is largely used in molecular biology, it was first defined from a morphological perspective at the beginning of the century [see Sattler (1988) for a review of this concept in developmental morphology]. Before the recent development of molecular genetics this concept was used in a developmental morphological sense without specific reference to homeotic genes (Sattler 1988). Within the framework of the phenomenon of homeosis, the inflorescence of certain members of the Araceae constitutes an interesting system in which to analyse the shift between organs of different types at the ontogenetic and morphological levels (e.g. Barabé and Lacroix, 2000; Barabé et al., 2000).

In the Araceae, the genus Calla has been used as an example of homeosis in flowering plants (Lehman and Sattler, 1992). In this specific situation, it was hypothesized by Barabé and Labrecque (1983) and Lehman and Sattler (1992) that tepals were replaced by stamens during the course of evolution. However, these authors did not observe the initiation of tepals and stamens on the same whorl on individual flowers. Recently, the flower of Philodendron has been used as a model to analyse theoretical aspects of morphogenesis, such as the phenomena of homeosis, morphogenetic gradient, and positional information (Barabé and Bertrand, 1996; Barabé and Lacroix, 2000; Barabé et al., 2000).

In inflorescences of Philodendron, as is the case for other members of the Aroideae sensu Mayo et al. (1997), female flowers are located in the lower portion and male flowers in the upper portion. The presence of atypical bisexual flowers (ABFs) called ‘monströse Blüten’ in the intermediate zone of the inflorescence of genera with unisexual flowers (e.g. Philodendron and Homalomena) was first noted by Engler and Krause (1912, p. 16). Atypical bisexual flowers in the genus Philodendron were described by Mayo (1986, Fig. 393) on the inflorescences of P. × evansii (subgen. Meconostigma), and studied in more detail in P. acutatum (Boubes and Barabé, 1996). This phenomenon has been well documented in the genera Cercestis (Barabé and Bertrand, 1996) and Philodendron (Boubes and Barabé, 1996; Barabé and Lacroix, 1999; Barabé et al., 2000). Recently, based on the developmental morphology of ABFs of Philodendron fragrantissimum, P. grandifolium, P. megalophyllum, P. melinonii and P. solimoesense (Barabé et al., 1997, 2000; Barabé and Lacroix, 1999, 2000), it was reported that staminodes and carpels are initiated on the same whorl. We consider this an exceptional case of homeosis.

Teratological examples of intermediates between stamens/carpels inserted on the same whorl have been reported in a variety of taxa. Floral organs with both staminal and carpellar characteristics in specific whorls have been reported in Tulipa, Tofeldia and Sempervivum by Guédès (1979). In Philodendron, however, staminodes and fully functional carpels are observed on the same whorl in ABFs. To our knowledge, this phenomenon appears to be unique to Philodendron in the angiosperms. Homeotic transformations involving stamens and carpels have also been reported by Sawhney (1992). He showed that the development of stamen primordia in stamenless‐2 mutants of tomato can be modified with certain combinations of specific plant growth regulators and different temperature treatments to develop into carpels. Similarly, Lecocq (1972) reported cases of anomalous flowers in Begonia tuberhybrida where intermediate organs with characteristics of both stamens and carpels are found in a continuous transition from typical stamens to carpels. The arabidopsis epi‐mutant car recently described by Rohde et al. (1999) has more than two carpels, more than six stamens and chimeric organs that represent, as the authors state (p. 963), ‘fusions of stamen and carpel’. Other current morphological studies of homeosis involve mainly perianth–stamen mutations (Barabé and Labrecque, 1983; Kirchoff, 1991; Lehman and Sattler, 1992, 1993, 1996; Tucker, 1992; MacIntyre and Lacroix, 1996).

The developmental morphology of the flower of Philodendron provides a new opportunity to analyse quantitatively the phenomenon of homeosis. Previous studies based on a qualitative analysis have shown a high degree of variability in the mode of expression of carpel/stamen in homeotic bisexual flowers. This was not only observed in species belonging to different subgenera but also in individuals of the same species. To date, no quantitative analyses of these variations have been performed to determine whether there are some regularities in the homeotic shift in ABFs. For example, is there a relationship between the number of staminodes and the number of carpels inserted on the same whorl? How many carpels are replaced by a staminode in a given species? Is there a one‐to‐one organ replacement? To answer these questions, a system where the different types of flowers are contiguous on the same surface, as is the case in inflorescences of Philodendron, proved to be very useful. We address these questions by analysing the quantitative relationships between different types of atypical bisexual flowers in P. squamiferum and P. pedatum. However, to determine whether there is a quantitative relationship, a strong descriptive framework is needed for reference. Thus, in the first instance we present developmental data to show the qualitative variation in floral structure in these two species. These new morphological and ontogenetic characters related to the development of flowers in P. squamiferum and P. pedatum will also complement anatomical studies of subgenus Philodendron species (Barahona Carvajal, 1977; Mayo, 1986, 1989), and add to our present knowledge of floral types within this group.

Previous work by Barabé and Lacroix (2001) has shown that there are two ways to define homeotic changes in ABFs. Bisexual atypical flowers can be considered as a modified female flower or a modified sterile male flower. The qualitative determination of the number of carpels and staminodes involved in the homeotic transformation will not necessarily be the same in both cases. This can lead to a problem of morphological interpretation especially if the homeotic shift does not involve a one‐to‐one organ replacement. We intend to show that the determination of a quantitative relationship between the mean number of staminodes and carpels in ABFs will offer a precise conceptual framework to analyse this problem.

The general goals of this study are to: (1) compare the development of flowers of P. squamiferum and P. pedatum with that of other previously studied members of the same subgenus; (2) characterize the range of homeotic transformation in these species; and (3) verify whether there is a quantitative relationship between the number of carpels and the number of staminodes in the homeotic transformation occuring in the ABFs.

MATERIALS AND METHODS

Plant material

Philodendronsquamiferum Poepping and Philodendron pedatum (Hooker) Kunth belong to the subgenus Philodendron (Croat, 1997). Specimens used for this study were collected in French Guiana (Petit‐Saut road) in May and November 1997. Voucher specimens have been deposited at the Herbier Marie‐Victorin (MT): Barabé 34, Barabé 35, Barabé 59. Specimens of P. pedatum growing at the Montreal Botanical Garden were also collected in April 2000 (registration numbers: 191‐97; 2043‐1997). Inflorescences at various stages of development were dissected under a stereomicroscope to expose the spadix, and fixed in formalin–acetic acid–alcohol (1 : 1 : 9 by volume), and later transferred and stored in 70 % ethanol.

Microscopy

Thirty‐six inflorescences of P. pedatum and 32 inflorescences of P. squamiferum were dehydrated in a graded ethanol series to absolute ethanol. They were then dried in a LADD model 28000 critical point dryer using CO₂ as a transitional fluid, mounted on metal stubs and grounded with conductive silver paint. Specimens were sputter‐coated with gold/palladium to approx. 30 nm using a Denton Vacuum Desk II sputter coater (Moorestown, NT, USA), and viewed with a Cambridge S604 scanning electron microscope (SEM) (Cambridge, UK) with digital imaging capabilities (SEMICAPS^®) (Santa Clara, CA, USA).

Quantitative analysis

The following parameters were measured on five inflorescences of P. squamiferum and six inflorescences of P. pedatum ranging from 3 to 12 mm in length: number of stamens on five male flowers; number of staminodes on five sterile male flowers; diameter of three carpel primordia; diameter of three stamen primordia; diameter of three staminode primordia; and number of staminodes and carpels on all atypical bisexuals flowers available on an inflorescence. Inflorescence specimens used for the SEM were mounted on their long side on the stub. Consequently, measurements could only be made on one side of the inflorescence.

Measurements were used for regression or partial correlation analyses. In some situations, partial correlations were calculated for the following reasons. If X and Y are correlated in a linear regression, it could indicate a real or an apparent connection. This occurs when two variables (x and y), although independent, are linked to a third variable (z). The partial correlation, by analysing the concomitant variations of X and Y, when the third variable is kept constant, allows recognition of this hidden link.

All analyses were performed using STATGRAPHICS (Graphic Software Systems Inc., Version 4·0; Rockville, VA, USA) or STATISTICA (Statsoft Inc., Version 5·5; Hamburg, Germany) to determine whether there was a significant correlation between the number of floral parts. Ultimately these data can be used to show the quantitative relationships that can exist between female flowers, male flowers, sterile male flowers and ABFs.

It is important to note that in a sample of n flowers there are at least two interesting sources of variation: between flowers on each inflorescence and between inflorescences in the sample. If variance between inflorescences is not taken into account in each sample, the extent of the confidence interval of the mean increases automatically. Consequently, the variance tied to the inflorescences does not invalidate the interpretation when the statistical test indicates a significant difference between the means of two samples. In the other case, of course, an ANOVA is needed.

RESULTS

Developmental morphology

Morphology of mature flowers.

The length of the mature spadices of P. pedatum ranges from 16 to 18 cm, and that of P. squamiferum from 7 to 10 cm. Both staminate and pistillate flowers have no perianth. In both species, staminate flowers occupy the upper portion of the inflorescence and make up approx. 60 % of the total length of the inflorescence, whereas female flowers are located on the lower portion and occupy approx. 30 % of the total length. Between the male and the female portions of the inflorescence, there is an intermediate zone (approx. 10 % of the total length) consisting of sterile males flowers and ABFs.

In both species, modified stomata (water pores) in the sense of Vogel (1977) are found on the surface of the apical portion of the stamens (Fig. 1A and B) and staminodes on male and sterile male flowers, respectively. The epidermis of stamens and staminodes consists of pegged and ridged cells (Fig. 1B). In the male zone, the mature spadix is lined with longitudinal resiniferous canals located at the base of the stamens (Fig. 1C–E). In cross‐section, they form a ring of tubes inserted at the periphery of aerenchymatous tissue. In nearly mature inflorescences, young secretory cells enclose the cavity of the canals (Fig. 1D and E), which become filled with resin during later stages of development (Fig. 1C). This type of resin canal with a papillose epithelial layer corresponds to the Philodendron smithi type, sensu Mayo (1986, Figs 7·3·3).

Fig. 1. Mature flowers of Philodendron squamiferum and P. pedatum. A, Top view of surface of stamens (S) of P. squamiferum. Bar = 300 µm. B, Close‐up of staminal surface showing the pegged and striated epidermal cells. Note the presence of a stomatum (arrowhead). Bar = 150 µm. C, Side view of stamens (S) of P. squamiferum and resin canals (arrowheads) at the periphery of the inflorescence axis. Bar = 300 µm. D, Similar view of stamens (S) and resin canals (arrowheads) in P. pedatum. Bar = 300 µm. E, Close‐up of a resin canal of P. pedatum showing cellular extensions (arrowheads). Bar = 75 µm. F, Intermediate zone of the inflorescence of P. squamiferum showing the female flowers (F) and staminodes (St). Bar = 300 µm. G, Close‐up of a mature female flower of P. pedatum showing stigmatic surface. Bar = 300 µm.

The bisexual flowers generally consist of carpels and staminodes inserted on the same whorl. The portion of the bisexual flower facing the male zone consists of staminodes, and the portion facing the female zone consists generally of an incomplete gynoecium (Figs 1F, 6G and 7B). On mature flowers, it is difficult to determine whether the staminode(s) and carpels are inserted on the same whorl (cf. Fig. 1F). However, during early stages of development this phenomenon is visible (see section on floral development of ABFs).

Fig. 6. Development of sterile male flowers in P. squamiferum and P. pedatum. A, Early stage of development of the intermediate portion of the inflorescence of P. squamiferum showing male (M), sterile male (SM) and female (F) floral primordia. Arrows highlight the continuity of parastichies across the different floral zones. Bar = 150 µm. B, Early stage of formation of staminodes (arrowheads) in P. squamiferum at the periphery of the floral meristem. Bar = 150 µm. C, Junction between the sterile male floral zone and the female zone in P. squamiferum. Note the presence of an atypical bisexual flower (ABF). Bar = 150 µm. D, Early stage of initiation of floral primordia in the intermediate zone on the inflorescence of P. pedatum. Bar = 75 µm. E, Initiation of staminodes (arrowheads) at the periphery of the floral meristem in P. pedatum. Bar = 75 µm. F, Later stage of development of staminodes in P. pedatum showing a more compact arrangement of floral organs. Bar = 300 µm. G, Intermediate zone in P. pedatum showing the presence of two atypical bisexual flowers (asterisks). Bar = 300 µm.

Fig. 7. Variety of atypical bisexual floral types in P. squamiferum and P. pedatum. A, General view of intermediate zone in P. squamiferum showing three atypical bisexual flowers (asterisks). Bar = 150 µm. B, Three atypical bisexual flowers in P. pedatum consisting of a gynoecial portion (G) and staminodes (arrowheads). Bar = 300 µm. C, Single atypical bisexual flower consisting of one carpel (C) and four staminodes (arrowheads). Bar = 75 µm. D, Atypical bisexual flowers P. pedatum with an open locule (arrowhead) and associated staminode (St?). Bar = 150 µm. E, Two atypical bisexual flowers in P. squamiferum. Flower A consists of one (possibly two) staminodes (St, St?), an open carpel (arrow), and six carpels (C). Flower B has three carpels (C) and three (possibly four) staminodes (St, St?). Bar = 150 µm. F, Atypical bisexual flower in P. pedatum with two carpels (C) and four staminodes (St), one of which is continuous with the gynoecial portion (arrowhead). Bar = 150 µm. G, Variation in extent of staminode continuity with the gynoecial portion (arrowhead) in P. pedatum. Bar = 300 µm. H, Example of atypical bisexual flower of P. pedatum with a continuous (arrowhead) and separate (St) staminodes. Bar = 150 µm.

Unlike bisexual flowers, mature female flowers have a prominent stigmatic surface (Fig. 1G).

Inflorescence and floral development.

The inflorescence primordium of P. squamiferum is more or less cylindrical in shape during early stages of initiation (Fig. 2A). The different types of flowers are initiated acropetally along the axis of the inflorescence. Pistillate flowers develop on the basal portion of the inflorescence and staminate flowers develop on the terminal portion. At this stage of development, primordia of the intermediate zone are not as clearly outlined as those of the other two zones (Figs 2A and 6A). Floral primordia of the female zone cover approx. one‐third of the inflorescence at this very early stage (Fig. 2A). Once all the floral primordia have been initiated, the different types of flowers are all approximately the same size (Fig. 6A). It is interesting to note that there is no discontinuity in the phyllotactic pattern of the flowers between the different zones of the inflorescence. Pistillate flowers, sterile male flowers, atypical bisexual flowers and staminate flowers are inserted on the same contact parastichies (Figs 2A and 6A, long arrows).

Fig. 2. Stages of development of the male flowers of P. squamiferum. A, Young inflorescence showing an early stage of development of male (M), sterile male (MS), and female (F) floral primordia. Bar = 300 µm. B, Close‐up of early stage of development of male (M) floral primordia. Bar = 75 µm. C, Initiation of stamens (arrowheads) on male floral primordia. Bar = 75 µm. D, Later stage of male floral development showing more prominent stamen primordia (arrowheads). Bar = 75 µm. E, Later stage of development of stamens showing early stages of theca formation (arrowheads). Bar = 75 µm. F, Nearly mature stamens (S). Bar = 150 µm.

Stamen primordia are initiated simultaneously on the periphery of more or less circular floral primordia (Fig. 2B, arrowheads) on one whorl in both P. squamiferum (Fig. 2C and D) and P. pedatum (Fig. 3A). There is a mean of 3·9 stamens (three to five) per flower (Table 1 and Fig. 2D) in P. squamiferum and 6·0 (four to eight) in P. pedatum (Table 1 and Fig. 3B and C). During later stages of development, floral primordia come into contact with each other (Figs 2D and E and 3B and C). In addition, the size of stamens increases to the point that they eventually occupy all the available space between flowers (Figs 2E and F and 3C and D).

Fig. 3. Stages of development of the male flowers of P. pedatum. A, Early stage of initiation of stamens (arrowheads) at the periphery of the floral meristem. Bar = 75 µm. B, Stamens at the stage of theca formation (arrowheads). Bar = 150 µm. C, Later stage of development of male flowers showing a more compact arrangement of stamens (S). Bar = 150 µm. D, Nearly mature stamens (S). Bar = 150 µm.

Table 1.

Mean number of appendages in different types of flowers of P. squamiferum and P. pedatum

	P. squamiferum	P. pedatum
Male flowers	3·87 (n = 30) (3·67–4·06)	6·00 (n = 20) (5·41–6·59)
Sterile male flowers	5·43 (n = 30) (5·08–5·78)	5·60 (n = 20) (5·16–6·04)
Female flowers	8·00 (n = 30) (7·67–8·33)	8·76 (n = 25) (8·30–9·21)
Atypical bisexual flowers	6·76 (n = 17) (6·30–7·23) (stamiodes + carpels)	6·58 (n = 12) (5·71–7·46) (stamiodes + carpels)
Carpels in ABF	3·41 (n = 17) (2·56–4·27)	4·50 (n = 12) (3·22–5·79)
Staminodes in ABF	3·35 (n = 17) (2·65–4·06)	2·08 (n = 12) (1·25–2·92)

Confidence interval at 95 % in parentheses. ABF, Atypical bisexual flowers.

During early stages of development, female floral primordia have a hemispherical shape (Fig. 4A). Carpel primordia are initiated on the periphery of the floral primordia in both species (Fig. 4B). Carpel primordia on individual flowers have an inward facing horseshoe shape (Figs 4C and 5A). In some situations two adjacent floral primordia can become concrescent (Fig. 5A). During later stages of development, the entire ovary wall of typical flowers is formed by the concrescence of the walls of adjacent carpels (Fig. 5B). In P. squamiferum, there are, on average, eight (seven to ten) carpels per flower (Table 1 and Fig. 4C), compared with a mean of 8·8 carpels per gynoecium in P. pedatum, excluding the concrescent flowers (Table 1 and Fig. 5A and B). The small holes visible on the periphery of floral primordia during later stages of development represent the upper portion of the stylar canals (Figs 4D and 5C). Even though the carpels are concrescent, individual stylar canals are found in the mature ovary up to a level directly below the stigma. The formation of the compitum results in one opening during the final stages of floral development (Fig 4D and 5D). Papillae form on the stigmatic surface when the ovary is nearly mature in both species (Figs 1G and 5D).

Fig. 4. Stages of development of the female flowers of P. squamiferum. A, Early stage of initiation of the female floral primordia (F). Bar = 75 µm. B, Early stage of formation of carpel primordia (arrowheads) at the periphery of the floral meristem. Bar = 75 µm. C, Later of stage of development showing the formation of locular cavities (arrowheads). Bar = 75 µm. D, Female flowers prior to the formation of stigmatic surface. Note the formation of the stylar compitum (arrowheads). Bar = 150 µm.

Fig. 5. Stages of development of the female flowers of P. pedatum. A, Formation of locular cavities (arrowheads). Note the presence of flowers with supernumerary number of carpels (asterisks). Bar = 150 µm. B, Later stage of development showing a more prominent ovary wall. Bar = 150 µm. C, Close‐up of female flowers before the formation of a stigmatic surface showing stylar cavities surrounding the raised central portion of the gynoecium (arrowheads). Bar = 150 µm. D, Nearly mature female flower with a well‐developed stigmatic surface. Bar = 300 µm.

In P. squamiferum and P. pedatum, sterile male flowers and ABFs of the intermediate zone occupy approximately 10–15 % of the total length of the inflorescence. These flowers form a transition zone between typical male and female flowers. During early stages of development, the primordia of sterile male flowers and ABFs are approximately the same shape as staminate floral primordia and female floral primordia (Figs 2B, 4A and 6D) in both species.

The primordia of the staminodes on the sterile male flowers are initiated on the periphery of the floral primordium (Fig. 6E). In P. squamiferum, there are, on average, 5·4 staminodes (four to nine) per flower (Table 1 and Fig. 6B), compared with 5·6 (four to seven) primordia per flower in P. pedatum (Table 1 and Fig. 6E, F).

Although the floral organs of the sterile male flowers and ABFs are initiated later than those of the pistillate flowers, their relative rate of growth is faster than that of the other types of flowers due to the development and expansion of the staminodes. As a result, sterile male flowers and ABFs are larger than female flowers at maturity (Figs 1F and 6C).

ABFs form a more or less continuous single row on the inflorescence and are located directly below the sterile male flowers (Fig. 7A and B). Primordia of the floral organs are initiated on the periphery of a discoid floral primordium (Fig. 6E), and their nature, number and form vary considerably. In both species, the types of floral organs produced on bisexual flowers are correlated to their proximity to the other floral zones. Female organs of ABFs tend to be initiated on the side of the flower adjacent to the female zone and male organs are initiated on the side of the flower closer to the sterile male floral zone (Fig. 7A, B, D and H). In P. squamiferum, the number of carpels in ABFs ranges from one (Fig. 7A) to six (Fig. 7E flower A), as does the number of staminodes (Fig. 7A), with mean values of 3·41 and 3·35, respectively (Table 1). The mean total number of floral organs (carpels + staminodes) per ABF is 6·76 (Table 1); it is never less than five or greater than eight. In the ABFs of P. pedatum, the mean total number of carpels ranges from one (Fig. 7C) to seven (Fig. 7E), and the number of staminodes from one (Fig. 7H) to five (Fig. 7B) and (Table 1), with a mean of 4·5 and 2·08, respectively (Table 1). The mean total number of floral organs (carpels + staminodes) per ABF is 6·58 (never less than five or greater than nine). Although it is easy to identify floral appendages of specific flowers during early stages of development, it is often difficult to determine which staminodes belong to which flowers in later stages of development in ABFs (Figs 6G and 7D and E).

During early stages of development, the surface area of individual ABFs of P. squamiferum is, on average, greater than that of female flowers regardless of the number of carpels and staminodes involved. This is visible, for example, in Fig. 7A and E, where the ABFs are larger than the surrounding female flowers.

The staminodes and the adjacent carpels in ABFs form a continuous ring and are inserted on a single whorl (Fig. 7A and C). In many ABFs, however, there is an incomplete separation between staminodes and carpels. Figure 7B, F and G shows ABFs with staminodes united to the gynoecium by a residual portion of the ovary wall. In some flowers, one open carpel is inserted on the same whorl as typical carpels (Fig. 7D and E). In Fig. 7D, the aberrant structure (arrow) is clearly identifiable as an open carpel. However, in some ABFs (Fig. 7E and H, arrows) it is difficult to determine whether the aberrant structure corresponds to an open carpel or a staminode. On ABF (A) in Fig. 7E, there are six closed carpels, one staminode, and one aberrant structure corresponding to an open carpel. From a morphological point of view, however, this aberrant structure (arrow in Fig. 7E) could also represent an intermediate structure between an aborted open carpel and a typical stamen. Figure 7H shows such an intermediate structure (arrow) between an open carpel and a typical staminode. These morphological observations provide strong evidence that staminodes and carpels are inserted on the same whorl in ABFs.

Quantitative results

P. squamiferum.

In ABFs, the mean number of appendages (staminodes + carpels) is less than the number of carpels in female flowers and larger than the number of staminodes in sterile male flowers. However, there is no significant difference between the mean number of staminodes (3·35) and carpels (3·41) per flower in ABFs (Table 1).

There is a strong negative correlation (r² = 0·704) between the mean number of staminodes and the mean number of carpels in ABFs (Fig. 8). The linear regression in question is described by the equation y = 6·8 – 1·02x, where x represents the number of staminodes and y the number of carpels. The value of the slope (–1) indicates that, on average, one staminode replaces one carpel. In fact, the equation shows that the mean total number of appendages (staminodes + carpels) must be equal to 6·83, a value that is not significantly different from the mean number of total appendages observed in ABFs (Table 1). Therefore, there appears to be a one‐to‐one organ replacement in homeotic flowers of P. squamiferum.

Fig. 8.P. squamiferum. Regression of the number of carpels (y‐axis) in relation to the number of staminodes (x‐axis) in ABFs. n = 17.

P. pedatum.

In ABFs, the mean number of appendages (staminodes + carpels) is inferior to the number of carpels in female flowers and superior to the number of staminodes in sterile male flowers. The mean number of staminodes (2·08) and the mean number carpels (4·50) per flower are significantly different in ABFs (Table 1).

The number of samples of P. pedatum available for statistical analysis was low: the linear regression takes only 12 pairs of values into account. These correlations can therefore be interpreted as indicators of general tendencies. As in P. squamiferum, there is also a negative significant correlation (r² = 0·543) between the mean number of staminodes and the mean number of carpels in ABFs (Fig. 9), corresponding to regression equation y = 6·87 – 1·14x, where y represents the number of carpels and x the number of staminodes. The value of the slope (–1·14) is not significantly different from –1. This indicates that on average, as in P. squamiferum, one staminode replaces one carpel. In fact, the regression equation shows that the mean total number of appendage (staminodes + carpels) must be equal to 6·9, a value that is not significantly different from the mean number of total appendages observed in ABFs (Table 1). Overall, there is, on average, a one‐to‐one organ replacement in homeotic flowers in both P. pedatum and P. squamiferum.

Fig. 9.P. pedatum. Regression of the number of carpels (y‐axis) in relation to the number of staminodes (x‐axis) in ABFs. n = 12.

DISCUSSION

Atypical bisexual flowers

Although the pistillate portion of bisexual flowers is fertile, the staminate portion consists of sterile stamens referred to as staminodes. From a functional reproductive point of view, these flowers remain unisexual. On the other hand, from a developmental point of view, these flowers can be considered bisexual because pistillate and staminate primordia are initiated on the same floral primordium, even if the staminate primordia develop into staminodes at a later stage.

Even though the developmental morphology of ABFs in P. squamiferum and P. pedatum is similar to that of the other species studied to date, the ABFs of the species used in this study have a greater diversity in gynoecial morphology. Nonetheless, in both species, the incomplete separation of staminodes from the gynoecial portion of the whorl (Fig. 6F) is strong evidence that the staminodes and carpels belong to the same whorl.

The presence of a morphogenetic gradient in the intermediate zone of the inflorescence as described for P. fragrantissimum (Barabé et al., 2000), P. grandifolium (Barabé and Lacroix, 2001), P. melinonii (Barabé and Lacroix, 2000) and P. solimoesence (Barabé and Lacroix, 1999) applies equally to P. pedatum and P. squamiferum. This gradient corresponds to a qualitative variation in the nature of the appendages that are initiated. The continuous morphological transition in sexuality between female flowers, ABFs, sterile male flowers and male flowers is also visible on individual bisexual flowers where the portion facing the female zone of the inflorescence displays female characteristics (carpels) and the portion facing the male zone displays male characteristics (staminodes). All ABFs of Philodendron species studied to date by our group are the result of a unique manifestation at the level of the inflorescence of a qualitative gradient affecting the nature of the appendages that are initiated in single flowers within that gradient. This idea of a morphogenetic gradient has been used (within the framework of the theory of positional information) to explain the development of ABFs on the inflorescence of many species of Philodendron (Barabé and Bertrand, 1996; Boubes and Barabé, 1996; Barabé et al., 2000).

In P. squamiferum and P. pedatum, a phenomenon was observed that has not yet been well documented in other species of Philodendron: the incomplete separation of carpel and staminode primordia inserted on the same whorl (Figs 6G and 7B and D–H). A case of incomplete separation of the gynoecium and staminodes, similar to that represented in Figs 6G and 7H, was reported in Philodendron melinonii (see Barabé and Lacroix, 2000; Fig. 6D and E). In specific situations, in P. squamiferum (Fig. 7E), normal carpels (arrowheads), an appendage morphologically intermediate between a typical carpel and a staminode (arrow), and a staminode are inserted on the same whorl. The intermediate nature of some floral organs indicates that they appear to be under the physiological influence of both the neighbouring female and male organs. In rare situations, as in Fig. 7E (arrow), one can observe what may be interpreted as an open carpel. In other cases, typical staminodes that are continuous with the gynoecial portion are also observed (Fig. 7B and F–H). In Figs 7B and H, for example, ABFs consist of typical carpels, a free staminode and a staminode that is continuous with the gynoecial portion, all of which are inserted on the same whorl.

Homeosis

As in other species of Philodendron that have been studied, ABFs of P. squamiferum and P. pedatum constitute a case of homeosis where there is a shift on the same whorl between carpels and staminodes. In P. squamiferum and P. pedatum there is a significant correlation between the number of carpels and the number of staminodes in ABFs. This indicates that there is a certain degree of regularity in the number of organs involved in the homeotic transformation occuring in ABFs. In both species, the slope of the linear regression of those two parameters is not significantly different from –1. In both species one carpel is, on average, replaced by one staminode. However, the mean number of carpels in ABFs of P. pedatum is twice that of staminodes. In this case, even though a carpel can be replaced by a single staminode, few carpels can actually be replaced. This appears to indicate that the mean number of different types of floral appendages on an ABF and the number of organs involved in a homeotic transformation are two independent phenomena. This quantitative result is in accordance with recent molecular studies that have shown that organ identity and organ number per whorl are indeed controlled by two different groups of genes (e.g. Running et al., 1998). In ABFs it is plausible to think that the potentiality to form carpels is greater than that to form staminodes.

The mean total number of appendages in ABFs is intermediate between that of female flowers and sterile male flowers. This implies that ABFs are influenced by both types of flowers located in two different zones of the inflorescence. In addition, the total number of appendages of ABFs in both species never exceeds the total number of carpels in female flowers. Thus, the number of appendages in female flowers may be imposing a constraint on the maximum total number of appendages (carpels or staminodes) that can develop in ABFs. However, the total number of staminodes in ABFs may exceed that of staminodes in sterile male flowers. In other words, at the global level of the inflorescence, the number of appendages in typical normal flowers of both sexes acts as boundary conditions for developmental dynamics involved in the formation of ABFs.

In a previous study (Barabé and Lacroix, 2000), it was hypothesized that for a given species the number of carpels replaced by an individual staminode (N) in atypical bisexual flowers is proportional to the ratio between the number of carpels (C) and number of staminodes (S) in unisexual flowers of that species (i.e. N ∼ C/S). This interpretation was based on a qualitative analysis. At this point, the sample used here is too small to draw a firm conclusion. However, a partial correlation analysis using the inflorescence as the basal unit (n = 6), indicates that there is no link between the number of staminodes and carpels in bisexual flowers and the number of appendages in other types of flowers (Barabé, Lacroix and Jeune, unpubl. res.). Therefore the hypothesis (N ∼ C/S) is not currently supported by these data. These preliminary results will have to be corroborated by the analysis of a greater number of samples.

On the nature of the carpel

In P. squamiferum and P. pedatum, the quantitative analysis indicates that on average one staminode replaces one carpel. When we refer to a shift between carpels and staminodes, the carpel is considered as a functional unit. In other words, the carpel is viewed as a structure enclosing the locules of the ovary, and consequently the number of carpels is equivalent to the number of locules. However, from a morphological point of view, it is difficult to determine whether the staminode in ABFs is homologous to a full carpel in the classical sense or to the portion of the ovary wall separating two carpels. Based on the quantitative results, staminodes of ABFs, as represented for example in Fig. 7A, can be interpreted as homologous to a carpel, while those represented in the ABF in Fig. 7C are homologous to two carpels. In contrast, in other types of ABFs such as those in Fig. 7E, there is a structure (flower A, arrow) with a form that is intermediate between a normal carpel and a typical staminode. This structure looks like an open carpel united to the gynoecium by a portion of the ovary wall adjacent to two carpels. The same interpretation can apply to a typical staminode that is united to the gynoecial portion of the flower. A similar phenomenon occurs in ABFs of P. pedatum (Fig. 7G and H, arrowheads), where the nature of the portion of tissue joining the staminodes to the gynoecium remains undetermined. Therefore there are two levels of aberrant floral structures in Philodendron. The first one is represented by the presence of ABFs, which are intermediates between typical female flowers and typical sterile male flowers. The second one is the presence of intermediate structures between typical carpels and typical staminodes on a single ABF. These particular structures developing in many ABFs seem to indicate a morphological continuum between typical carpels and staminodes.

With regards the morphological interpretation of the ABF, one can ask whether the bisexual flower is a modified female flower or a modified staminate sterile flower (Barabé and Lacroix, 2001). In P. squamiferum and P. pedatum, the quantitative determination of the number of carpels and staminodes involved on average in the homeotic change has shown that the homeotic shift involves a one‐to‐one organ replacement. Therefore, we can interpret ABFs as pistillate flowers where carpels have been replaced by an equal number of staminodes, or sterile male flowers where staminodes have been replaced by carpels. All ABFs of Philodendron could be considered incomplete staminodia whorls in the sense of Ronse Decraene and Smets (2001). However, as the total number of appendages in ABFs is significantly greater than the number of appendages in sterile male flowers, it is more plausible to interpret ABFs as modified female flowers.

The unique morphology of the inflorescence of P. squamiferum and P. pedatum allowed us to study a poorly known phenomenon in angiosperm flowers: quantitative homeotic transformations involving carpels and stamens on the same floral whorl. Based on the data available to date, the genus Philodendron presents us with a wide diversity of structure/function relationships relating to mechanisms of homeosis. Since the study of homeosis beyond arabidopsis and Anthirrinum has been conducted mainly at the morphological level, we plan to complement our work on Philodendron by characterizing the morphogenetic gradients at the physiological and molecular level in the near future.

ACKNOWLEDGEMENTS

This paper was mostly written during a visit by D.B. to the Laboratoire de Cytologie Expérimentale et Morphogenèse Végétale (Université Pierre et Marie Curie, Paris) in 2001. The first author would like to thank Pr Dominique Chriqui for her support, and the staff of the Laboratoire Environmental de Petit Saut (French Guiana) for technical support; also Dr Jean‐Jacques de Granville for permission to work at the Herbier de Guyane (ORSTOM, Cayenne), and Dr Alain Dejean (May 97) and Mr Manfred Korel (November 97) for their help in collecting specimens. This research was supported in part by individual operating grants from the Natural Sciences and Engineering Research Council of Canada to D.B. and C.L, and by a travel grant from the Bureau de la coopération internationale of the Université de Montréal to D.B.

Received: 25 March 2002; Returned for revision: 21 May 2002; Accepted: 15 July 2002 Published electronically: 2 October 2002