Contrasting lengths of Pelargonium floral nectar tubes result from late differences in rate and duration of growth

Abstract

Background and Aims

Much of morphological evolution in flowers has arisen from pollinator-mediated selection, often manifest as a match between the length of the pollinator’s proboscis and the depth of tubular corollas or spurs. We investigate development, growth and homology of the unique nectar tube of Pelargonium, frequently described as ‘a spur adnate to the pedicel’.

Methods

We focused on two species. The nectar tube of P. ionidiflorum is three times longer than that of P. odoratissimum. Light and scanning electron microscopy were carried out, and daily growth measurements were used to compare nectar tube development and vascular patterns.

Key Results

Nectar tubes in both species are initiated centripetally to the dorsal sepal in a space created by lateral displacement of two antepetalous stamens. The cavity deepens through subsequent intercalary growth of the receptacle that proceeds at the same rate in both species until tubes reach approx. 10 mm in length. Differences in final nectar tube lengths arise via an increase in the rate and duration of growth of the receptacle that begins just before anthesis (floral opening) and continues for several days past anthesis in P. ionidiflorum but does not occur in P. odoratissimum. Epidermal cells of the dorsal surface of the nectar tube in P. ionidiflorum are approx. 1.6 times longer than those in P. odoratissimum. Histological sections show no evidence that the nectar tube is a spur that became evolutionarily fused to the pedicel.

Conclusions

Nectar tubes in Pelargonium are localized cavities that form in the receptacle via intercalary growth. Differences in the rate and duration of growth just prior to and following anthesis underlie differences in final tube lengths. Because differences in cell lengths do not fully account for differences in nectar tube lengths, evolutionary diversification must involve changes in both cell cycle and cell expansion.

INTRODUCTION

Pollinator-mediated selection is recognized as a critical driver of the morphological evolution of flowers (Fenster et al., 2004; Johnson, 2010), and shifts in pollinators can contribute to clade diversification (Hodges et al., 2002; Fenster, et al., 2004; van der Niet and Johnson, 2012; van der Niet et al., 2014). Floral nectar tubes are often interpreted as key morphological traits associated with diversification because they enhance specialized relationships of flowers with distinct pollinators (Hodges, 1997). Tubular corollas and nectar spurs restrict nectar access to a narrow range of potential pollinators with mouthparts of appropriate length, and can increase the precision of pollen placement and deposition (Anderson et al., 2014; Armbruster et al., 2014), facilitating reproductive isolation. For multiple lineages, the presence of a floral or nectar tube is associated with greater species diversity when compared with related taxa (Hodges et al., 2002; van der Niet and Johnson, 2012) and there is often a close association between nectar tube length and proboscis length of the major pollinator [Aquilegia (Whittall and Hodges, 2007); Tritoniopsis (Anderson et al., 2014); Lithospermum (Cohen, 2012); Linaria (Fernandez-Mazuecos et al., 2013); Gymnadenia (Sun et al., 2015); but see also Vlasánková et al., 2017].

A remarkable example of lineage diversification in nectar tube length is the genus Pelargonium (Geraniaceae): nectar tube length ranges from <0.5 mm to >100 mm among the 280 species (Goldblatt and Manning, 2000; Bakker et al., 2005). Across the South African species, pollinators include bees, sunbirds, moths, different genera of long-proboscid flies and, rarely, butterflies (Vogel, 1954; Struck, 1997). Bakker et al. (2005) and others (Struck, 1997; Ringelberg, 2012) suggested that evolutionary shifts of pollinators and accompanying changes in nectar tube length played a critical role in diversification within Pelargonium, a lineage that constitutes the seventh largest radiation in the hyperdiverse Cape Floristic Region (CFR) of Southern Africa (Linder, 2003).

Evolutionary lability in Pelargonium nectar tube length suggests that the developmental changes underlying length differences could be relatively simple, for example through heterochronic shifts in growth rates or the timing of initiation (e.g. Alberch et al. 1979) of the nectar tube that do not involve reorganization of floral organs; however, remarkably little is known about the development of these nectar tubes. Here, we ask if differences in final nectar tube length in two closely related Pelargonium species arise through changes in the timing of initiation and/or the rate or duration of tube growth. In Aquilegia, a model system for studying spur evolution and development, longer spurs have a greater duration of growth, largely due to more extensive anisotropic cell elongation (Puzey et al., 2012). In Aquilegia, the nectar spur is initiated late in flower development as a freely projecting evagination of a petal. In contrast, the nectar tube of Pelargonium is developmentally integrated with the receptacle/pedicel (Vogel, 1998). Such pronounced integration with the receptacle suggests that tubes of different sizes are all likely to be initiated early in flower development because tube formation at later stages would disrupt multiple aspects of floral organization, including the initiation and arrangement of floral organs. Given the lability of nectar tube lengths across the genus, and the potential restriction of initiation to early developmental stages, we predict that variation in nectar tube length arises from differential rates or durations of growth rather than the timing of initiation. Thus, we compare early initiation and subsequent growth of nectar tubes of Pelargonium species that differ in nectar tube length.

Before we began our study, however, we discovered conflicting interpretations of the nature and homology of the nectar tube in Pelargonium. Nectar tubes are typically tubular corollas or spurs (tubular projections from perianth members). In contrast, the nectar tube of Pelargonium appears to be unique; nectar is not contained in a corolla tube or an extension of a perianth organ but instead in an internal channel closely associated with the receptacle and pedicel of the flower (Vogel, 1998; Jeiter et al., 2017). Nectaries in Pelargonium are frequently described as a spur adnate to the pedicel (Almouslem and Tilneybassett, 1989; Bakker et al., 2005; Bernadello, 2007; Ringelberg, 2012; Röschenbleck et al., 2014). Sauer (1933), and later Weberling (1989) described the nectar tubes as ‘axial spurs’, and considered them as ‘combined with the flower stalk’, but provided no explanation of how they develop. Several authors have interpreted the tubes as hypanthial (Miller, 1996; Struck, 1997), with little explanation. Japp (1909) translated by Link (1994) described the development of the nectar tube as a ‘stagnation of growth in a adaxial–extrastaminal area of the [receptacle]’, suggesting that the tube is receptacular. Labbe (1964) examined the anatomy of three species and also supported a receptacular interpretation. Therefore, our second goal was to explore further the homology of the nectar tube in Pelargonium.

No other genera in the Geraniaceae, including Hypseocharis, which is sister to the rest of the family, have nectar tubes, although in Monsonia nectar is produced in shallow concavities of the receptacle (Link, 1994; Jeiter et al., 2017). When comparisons with sister taxa are not informative, careful comparative anatomical and/or developmental analyses can offer insights into the nature of unusual structures. In some cases, ‘vascular conservatism’, i.e. the idea that the evolutionary history of fusion or shifts in position are preserved in the arrangement, degree of fusion and orientation of vascular bundles (Henslow, 1890; Moseley, 1967; Schmid, 1972), has been used to interpret evolutionary developmental history. While strict application of this concept has received criticism (Carlquist, 1969; Schmid, 1972), vascular patterns have been highly informative in some cases [e.g. elucidation of receptacular vs. appendicular epigynous ovaries (Kaplan, 1967)]. Because patterns of vasculature supplying the nectary of Pelargonium could provide evidence for evolutionary fusion of a freely projecting spur to the pedicel, we also examined transverse and longitudinal sections of mature flowers to document vascular patterns supplying the nectary.

MATERIALS AND METHODS

The study system

The genus Pelargonium L’Hér. is largely endemic to the Greater Cape Floristic Region (GCFR) of southern Africa. Within the CFR alone, Pelargonium is the third largest angiosperm genus (Goldblatt and Manning, 2002). Pelargonium exhibits remarkable morphological diversity of growth forms and leaf shape (van der Walt, 1977; van der Walt and Vorster, 1981, 1988; Jones, et al., 2009), and floral characters are variable even within sections (Struck, 1997; Röschenbleck et al., 2014).

We compare flower development in two closely related species, P. ionidiflorum and P. odoratissimum of section Reniformia (Bakker et al., 1998). Both are low-growing sub-shrubs. Native ranges of both species overlap, although P. odoratissimum is more broadly distributed in the western CFR. While we know of no reports of the specific pollinators of these species, they were chosen as exemplars of dramatic differences in tube length, ranging from 5–10 mm in P. odoratissimum (Goldblatt and Manning, 2000) to ‘ca. 35 mm’ in P. ionidiflorum (van der Walt and Vorster, 1981).

Scanning electron microscopy

Developing flower buds ranging in size from 0.5 to 4.0 mm were dissected to expose the nectar tube on the dorsal side. Dissected buds were fixed in FAA (Berlyn and Miksche, 1976) under refrigeration for 1 d. Fixed buds were dehydrated to 100 % ethanol, critical point dried (Model 931, Tousimis, Rockville, MD, USA), mounted onto scanning electron microscopy (SEM) stubs with double-sided tape, sputter coated with gold–palladium for <3 min (E5100, Polaron, acquired by Quorum Technologies, East Sussex, UK) and imaged with a scanning electron microscope (NOVA NanoSEM 450, FEI, Eindhoven, The Netherlands). Distractions were digitally removed from the backgrounds of the scanning electron micrographs using the stamp tool in Adobe Photoshop; contrast and brightness were adjusted in the sample image, but no other adjustments to images were made.

Nectar tube histology

Additional mature flowers of P. ionidiflorum were fixed in modified Karnovsky’s fixative (2.5 % gluteraldehyde, 2 % paraformaldehyde in PIPES buffer, pH 7.0), dehydrated and embedded in JB4 methacrylate (Polysciences, Warrington, PA, USA) and then sectioned at 5 µm. Sections were stained with aqueous 0.05 % Toluidine Blue O. Digital photographs were obtained with a Q-Imaging Micropublisher 3.3 camera (Qimaging, Burnaby, Canada) mounted on a Zeiss Axioskop (Zeiss, Thornwood, NY, USA) and acquired using Q-capture Pro 6.0 software.

Cell size analysis

For 6–7 fresh open flowers of each species, strips of epidermis from the dorsal surface of the nectar tube were carefully peeled away from the underlying tissue at the basal, mid and distal region of the tube. The epidermal strips were immediately mounted in distilled water and photographed using the system described above (see representative images in Supplementary Data Fig. S1). Epidermal peels were 3–7 mm in length, and 2–3 images of cells were acquired from different regions of each peel. The length and width of three of the longest cells per image were measured using ImageJ, for a total of 84 cells for P. ionidiflorum and 112 cells for P. odoratissimum.

We used mixed effects models that allow for different error variance at each level of nested analyses, e.g. cells within regions within plants, to test for differences between cell lengths and widths. Species and region of the tube (basal, mid and distal) were included as main effects, and plant plus image nested within plant were treated as random effects in various combinations using the ‘lme4’ R package (Bates et al., 2015). The best model was selected based on a comparison of Aikike information criterion (AIC) scores. Average differences in cell lengths and widths between species were compared with ‘aov’ (R base package; R Core Team, 2008) with plant as a random effect.

Growth and allometric analyses

To measure the rates and durations of perianth and nectar tube elongation, 20 buds per species (ten per plant, each on a different inflorescence) were tagged when they were 0.25–0.75 mm in length. Perianth and nectar tube length measurements were made daily between 12.00 and 13.00 h until the petals and the nectar tubes stopped growing. Perianths were measured from the base to the tip of the calyx and then from the base of the calyx to the tip of the corolla once it emerged beyond the calyx. Date of petal emergence and anthesis (flower opening) were recorded for each bud. One bud of P. odoratissimum showed unusually poor growth and failed to undergo anthesis, so it was later dropped from the analysis.

To compare growth trajectories, we used functional data analysis (FDA) (Ramsay and Silverman, 2005). Most statistical approaches for time series data either attempt to model the autocorrelation between measurements at each time point or fit relatively simple functions (e.g. exponential) to represent growth curves (e.g. Hunt, 1982; Kutner, et al., 2005). Functional data analysis estimates a single function that describes a continuous curve from growth data for each individual using a set of known basis functions, which are far more flexible and provide better fit to complex curves (Ramsay and Silverman, 2002). From the daily measurements of perianth and nectar tube lengths, growth data for each individual flower were converted into functions using the R package ‘fda’ (Ramsay et al., 2014). Because growth was non-cyclical and monotonic (i.e. no re-occurring patterns and always expanding), functions of growth curves were estimated with B-spline basis systems, positioning knots at each day. First and second derivatives were calculated from these growth curves in order to describe rates of change with time. Growth averages with ± 1 s.d. were estimated for tube and perianth by species for both growth and higher order functions. Calculations of 95 % confidence intervals for functions are not appropriate because characterizing the underlying functional distributions is an emerging technique in FDA. We chose to show the standard deviations to estimate the spread of the curves without making statements of statistical significance about the sample’s central location.

To quantify bivariate allometry between nectar tube and perianth elongation, slopes of log10-transformed daily measures of individual buds were calculated by standardized major axis regressions using the ‘smart’ package in R (Warton et al., 2012). Because the nectar tubes stopped growing before the perianth, slope calculations were restricted to the linear portion of the relationship, i.e. all measures up to 60 % of final tube length. Differences in allometric growth between species, using slopes for individuals as the response variable, were tested with ‘aov’ (R base package; R Core Team, 2008).

RESULTS

Mature flowers of P. ionidiflorum and P. odoratissimum are monosymmetric (Fig. 1A, C) with five sepals and five petals, two petals dorsal and three ventral. The two species differ in flower and nectar tube colour (Fig. 1A, C). Nectar tubes of P. ionidiflorum are nearly three times longer than those of P. odoratissimum (Fig. 1B, D; Table 1). Both species typically have seven fertile stamens that occur in two whorls; five are opposite the sepals (antesepalous) and the other two are opposite the dorsal petals (antepetalous) (Fig. 1E, F). The opening of the single nectar tube is between the dorsal sepal and the bases of the two antepetalous stamens (Fig. 1E, F, arrow). The proximal extent of the tube is marked externally by a small bump on the receptacle (Fig. 1A, C, arrows).

Fig. 1.

Mature flowers of Pelargonium ionidiflorum (A, B, E) and Pelargonium odoratissimum (C, D, F). (A and C) Front view of the corolla. (B and D) Lateral views; the nectar tube extends (internally) from the base of the dorsal sepal to the protrusion indicated by the arrow. The protrusion marks the location of the nectar-producing tissues. (E and F) Front views with sepals and petals removed. Arrows indicate the opening of the nectar tube. SP, antepetalous stamen, SS, antesepalous stamen. Scale bars = 5 mm.

Table 1.

Flower dimensions and growth dynamics of Pelargonium flowers

Species	Perianth length (mm)	Tube length (mm)	Days to petal emergence	Days to anthesis
P. ionidiflorum	17.85 ± 0.83	30.18 ± 2.32	17.15 ± 3.31	19.95 ± 2.64
P. odoratissimum	10.52 ± 0.6	10.05 ± 0.45	14.0 ± 1.49	15.5 ± 1.22

All values are given as the mean ± s.d.

Initiation and early development of the nectar tube

In all SEM images shown in Figs 2 and 3, sepals have been removed and buds are viewed dorsally, such that the full extent of the nectary is exposed. In 0.5–0.9 mm long buds of both species, all floral organs have been initiated (Fig. 2A, B). Petal primordia are small and delayed in development relative to sepals and stamens. Stamens show clear differentiation of anthers and filaments in both species. Dorsal views of buds reveal some expansion of the internode between the petals and antepetalous stamens (shown by a bracket in Fig. 2A, B), but little internode expansion between the two stamen whorls. Although the dorsal antesepalous stamen (dSS in Fig. 2A, B) is part of the outermost whorl of the androecium (Endress, 2010), it is displaced acropetally toward the gynoecium and flanked by two antepetalous stamens. The synorganization of these three stamens is associated with a relatively wide space between the two dorsal petals that anticipates the location of the nectar tube. At these early stages, however, a concavity indicating the initiation of the nectar tube is not evident in this region in either species (insets, Fig. 2A, B). Vertically aligned files of rectangular cells at the base of the dorsal antesepalous stamen (Fig. 2A, B insets) suggest the initiation of elongation in this region.

Fig. 2.

Scanning electron micrographs showing the similarity of early stages of floral bud development in Pelargonium ionidiflorum (A, C, E) and Pelargonium odoratissimum (B, D, F). All images show the dorsal side with the nectar tube exposed by removal of the dorsal sepal and dorsal nectar tube tissue. (A and B) 0.9 and 0.5 mm bud length, respectively, (C and D) 1.1 mm bud length, (E and F) 2 mm bud length. Inset, base of the receptacle; bracket, internode between petals and stamens; P, petal; SP, antepetalous stamens; SS, antesepalous stamens, dSS dorsal antesepalous stamen. Scale bars (A–F) = 100 μm.

Fig. 3.

Scanning electron micrographs showing continued similarity between species in slightly later stages of floral bud development of Pelargonium ionidiflorum (A, C) and Pelargonium odoratissimum (B, D). All images show the dorsal side with the nectar tube exposed. (A and B) 2.5 mm bud length, (C and D) 4 mm bud length. P, petal; SP, antepetalous stamens; SS, antesepalous stamens; dSS dorsal antesepalous stamen; C, carpel. Scale bars (A, B) = 500 µm; scale bars (C, D) = 1 mm.

In 1.1 mm buds, an incipient cavity is evident (Fig. 2C, D) in a localized region of the receptacle between the bases of the dorsal stamens and petals. Vertically aligned files of receptacle cells are evident subjacent to the insertion of the stamens (Fig. 2C, D insets) and the dorsal petals and sepal are borne on the rim of the nascent tube. For P. ionidiflorum, thecae of the anthers are becoming more clearly differentiated (Fig. 2C), and the antesepalous stamens have elongated slightly more than the antepetalous stamens.

In 2.0 mm buds (Fig. 2E, F), the developing nectar tube is longer compared with earlier stages and the vertically aligned cell files are longer, with each file having more cells, suggesting that continued growth occurs by ordered cell division (intercalary growth) (Fig. 2E, F insets). At the upper, distal limit of the tube, there is a distinct difference in shape between the rectangular cells of the tube and the polyhedral cells at the bases of the stamens for both species. Developmental stages of the petals and stamens appear similar for the two species in the buds sampled.

Subsequent development of buds from 2.5 to 4 mm in length is similar in the two species (Fig. 3). Nectar tubes continue to expand in length through intercalary growth of the receptacle between the position of insertion of the floral whorls and the base of the nectary. For both species, the three ventral antesepalous stamens are longer than the dorsal antesepalous and antepetalous stamens (Fig. 3A, B). In 4 mm buds of both species, the petals have grown larger relative to the antepetalous stamens (Fig. 3C, D) compared with younger stages. In P. ionidiflorum, petals are shorter than the antepetalous stamens (Fig. 3C), whereas in P. odoratissimum, the converse is true (Fig. 3D). Nectar tubes have elongated significantly, reaching almost 1 mm in length for samples of both species. Based on the flowers illustrated and SEM images of other flowers at similar stages but not shown, we conclude that the mode and timing of early stages of nectar tube development are essentially identical in the two species.

Perianth and nectar tube elongation rates

We quantified elongation of the perianth (sepals initially, then petals after emergence and anthesis) and nectar tubes by daily measurements of 20 buds of P. ionidiflorum and 19 buds of P. odoratissimum. Buds were 0.25–0.5 mm in length on day 1. Early perianth growth of the two species is similar (Fig. 4A). The rate of elongation in both species is somewhat biphasic (Fig. 4B). Initial rapid growth of the bud is followed by a decrease in the rate of expansion, and then an increase in growth rate associated with petal emergence and anthesis. In P. ionidiflorum, the growth rate is relatively uniform until approximately day 10, then declines until day 15 when the bud resumes more rapid elongation prior to petal emergence at day 17; mean maximum growth rates of >1.0 mm d^–1 are sustained from petal emergence on day 17 through day 24, 4 d beyond anthesis (Table 1). In contrast, the initial elongation rate of P. odoratissimum is slightly greater than that of P. ionidiflorum, on average, but declines more, reaching minimum rates at 10 d. Growth then accelerates to a maximum of approx. 1.0 mm d^–1 during corolla emergence at 14 d. This high rate is not sustained and begins declining even before anthesis at 16 d, ceasing completely by 17 d. Comparison of the two species shows that perianth growth of P. ionidiflorum buds continues well beyond the time that P. odoratissimum flowers reach anthesis and that longer petals of P. ionidiflorum primarily result from slightly higher maximum rates and, more importantly, a greater duration of rapid growth prior to anthesis.

Fig. 4.

Comparative growth and rate of growth curves of P. ionidiflorum and P. odoratissimum derived from functional data analysis (see text) based on daily measurements of bud lengths prior to anthesis and of total perianth length following anthesis. (A) Perianth growth. (B) Rate of perianth growth. (C) Nectar tube growth. (D) Rate of nectar tube growth. Red, P. ionidiflorum; blue, P. odoratissimum. Solid lines indicate mean curves, and shaded areas are standard deviations of the curves. Arrows, day of anthesis.

Elongation of nectar tubes of both species is measurable at day 5 of bud growth. Very early tube growth trajectories are similar (Fig. 4C, D), while mean curves for P. ionidiflorum suggest a slightly greater growth rate compared with P. odoratissimum, substantially overlapping ranges of variation show that these early differences may not be biologically meaningful. Growth curves of the two species diverge dramatically when tubes attain just under 10 mm average length (Fig. 4C, D). At this point, the mean growth rate of P. ionidiflorum nectar tubes increases dramatically to a maximum of 2.9 mm d^–1 on day 17, just prior to petal emergence for this species (Table 1). The rate of tube length growth declines after day 17, although overall growth continues until nectar tubes reach a mature length of 30.18 ± 2.32 mm at day 27, about 7 d after anthesis.

In contrast, mean tube growth rates in P. odoratissimum (Fig. 4C, D) increase only until about day 15, reaching a maximum of 1.0 mm d^–1. The period of maximum tube growth rates corresponds to the time of maximum petal elongation (compare Fig. 4B and D). Growth of the nectar tube ceases at the same time as that of the corolla (day 17), resulting in a mean mature tube length of 10.05 ± 0.45 mm, only one-third the length of the tube of P. ionidiflorum (Table 1). Thus, quantitative comparison of growth patterns indicates that the greater nectar tube length of P. ionidiflorum compared with P. odoratissimum results from both a longer duration and a greater rate of growth that begins just prior to corolla emergence and persists up to a week after anthesis.

Allometry of nectar tube elongation relative to perianth elongation

Although the flowers of P. odoratissimum are considerably smaller, their nectar tubes grow more rapidly relative to their perianth compared with P. ionidiflorum (Fig. 5) [P. odoratissimum slope = 5.77; P. ionidiflorum slope = 4.23; d.f. = 37, F_(1,37) = 13.86, P = 0.0007]. On an absolute time scale, however, the nectar tube and perianth of P. odoratissimu, cease elongation well before those of P. ionidiflorum (Figs 4 and 5B).

Fig. 5.

Comparative allometry of perianth and nectar tube growth in P. ionidiflorum and P. odoratissimum. (A) SMATR slopes of the linear portion of allometric growth of nectar tubes (i.e. 60 % final length) vs. perianth of individual buds in both species. Solid line, P. ionidiflorum; dashed line, P. odoratissimum. (B) Smoothed curves of 90 % of nectar tube length relative to perianth for both species. Red points, P. ionidiflorum; blue points, P. odoratissimum.

Cell sizes

Comparisons of epidermal cell lengths and widths from the dorsal surface of the nectar tube show that cells of P. ionidiflorum are 1.6 times longer, on average, than those in P. odoratissimum [x = 0.129 ± 0.04 mm (s.d.) vs. x = 0.079 ± 0.02 mm (s.d.), respectively, F_(1,190) = 183.3, P = 0.000; see boxplot in Supplementary Data Fig. S1]. Widths of P. ionidiflorum cells are greater than those of P. odoratissimum cells [x = 0.0141 ± 0.0029 mm (s.d.) and x = 0.0125 ± 0.0028, respectively; F_(1,190) = 21.73, P = 0.000]. The mixed effects model comparisons of cell length showed that the best model contained species as a fixed effect, and plant and image within plants as random effects. Including region (base, mid or top of the nectary) resulted in a higher log likelihood score (ΔAIC = 101.68).

Do vascular patterns provide evidence supporting evolutionary fusion?

If nectar tubes in Pelargonium evolved from adnation of a nectar spur to the pedicel, vascular patterns may provide evidence of this organ fusion. In Fig. 6, we diagram predicted vascular patterns for flowers with (Fig. 6A) a sepalar spur, (Fig. 6B) a spur fused to a pedicel and (Fig. 6C) a receptacular tube. Vascular conservatism (Fig. 6B) predicts that the vascular bundle supplying the dorsal sepal would take a convoluted path, extending up the pedicel to the point of sepal insertion before looping down the ventral side of the tube to supply nutrients into the dorsal sepal. Consequently, there would be two vascular bundles in cross-section on the ventral side of the spur that would have opposite orientations of xylem and phloem.

Fig. 6.

Predicted vascular patterns (in red) for flowers with (A) a sepalar spur, (B) a spur fused to a pedicel and (C) a receptacular tube. In (A), one of the petiolar vascular bundles branches at the receptacle to supply the spur and sepal. In (B), the path of the vascular bundle illustrated in (A) is retained, although the spur is adnate to the pedicel (a pattern consistent with predictions of vascular conservatism). Note that there would be two vascular bundles just ventral to the tube that would have opposite orientations of xylem and phloem. In (C), the petiolar vascular bundle branches below the nectary to supply the dorsal sepal, and there is a single vascular bundle just ventral to the tube.

Longitudinal sections through the nectar tube show a cylindrical cavity with densely staining secretory cells restricted to the base (Fig. 7A). The vasculature below the nectary shows acropetal divergence of a petiole vascular bundle to the dorsal side of the nectary tube (Fig. 7B) and no evidence that the bundle extends acropetally and back down as in Fig. 6B. Serial sections reveal that the region of the pedicel just proximal to the nectary contains six collateral vascular bundles arranged in a cylinder (Fig. 7C). Just below the level of the nectariferous tissue (Fig. 7D), there is a dorsal parenchymatous gap in the ring of bundles. Dorsal to the gap, a vascular bundle is oriented approximately perpendicular to the long axis of the tube, as indicated by the orientation of the tracheary elements (see arrow in Fig. 7D). A section 1 mm distal to the nectar tube base (Fig. 7E) shows the tube cavity with a single vascular bundle in the dorsal wall of the tube (cf. Fig. 6C). The receptacle contains a single ring of collateral vascular bundles (Fig. 7E arrows) with xylem oriented toward the centre of the axis.

Fig. 7.

Histological sections of the nectar tube of P. ionidiflorum following anthesis; all photos are oriented with the dorsal side of the tube to the left. (A) Longitudinal section through the base of the nectar tube and receptacle showing small, densely staining secretory cells at the base of the tube (indicated by an asterisk). (B) Lateral deviation of the dorsal vascular bundle of the petiole into the nectariferous tissue. This same bundle extends acropetally along the dorsal side of the nectar tube (see Fig. 6E). (C) Transverse section of the pedicel proximal to the nectar tube showing a cylinder of vascular bundles. (D) Transverse section through the dorsal region of the nectar tube base showing the lateral deviation of the vascular bundle in this region (arrow) and the absence of a dorsal vascular bundle within the cylinder of vascular bundles of the receptacle (curved arrow). (E) Transverse section 1 mm distal to the base of the nectar tube showing the cavity of the tube, the vascular bundle on the dorsal surface of the tube that is supplying the dorsal sepal (arrow) and a cylinder of vascular bundles in the receptacle, with a single bundle present in the dorsal position of this cylinder (curved arrow). Scale bar = 0.1 mm in all images.

DISCUSSION

Diversification in association with pollinators is often associated with changes in flower morphology, yet we know little about the developmental processes that underlie such changes in floral form. Here we show that for two species of Pelargonium that differ dramatically in nectar tube length, initiation and early development of nectar tubes are essentially identical; contrasting nectar tube lengths arise late in development and result from changes in rate and duration of tube elongation. Differences in rate are manifest primarily just before anthesis, and elongation of the nectar tube continues for up to 4 d following anthesis in the longer tubed species, P. ionidiflorum. We also show that neither the development of the nectar tube nor the patterns of vascular bundles provide support for an interpretation of the nectar tube as an evolutionary fusion of a perianth tube (i.e. adnate) to the receptacle, and that nectar tubes in Pelargonium are indeed receptacular (Japp, 1909; Labbe, 1964; Jeiter et al., 2017). This receptacular mode of nectar tube formation appears to be unique among flowering plants.

Nectar tube development and differences in length

Nectar tubes of P. ionidiflorum and P. odoratissimum first become apparent when buds are at the same developmental stage and roughly 1 mm in length. In both species, the pattern of dorsal petal and stamen initiation anticipates the location of the tube: the two dorsal petal primordia are spaced further apart than the ventral petals and, in combination with the acropetal displacement of the dorsal antesepalous stamen, create a broad area of the receptacle centripetal to the dorsal sepal where the tube will develop. The inception of the nectar tube is first recognized as a concavity of the receptacle in this localized region, with the tube eventually becoming embedded in the receptacle by intercalary growth. The presence of vertically aligned cell files lining the walls of the elongating tubes indicates that, at these early stages, growth of the receptacle results from regular, organized cell division. Nectar tube elongation is qualitatively similar in the two species up to 4 mm in bud length.

Similarities in tube growth also persist into later stages of flower development such that prior to 14 d, tube growth rates for the two species are similar. Quantitative analyses of nectar tube elongation demonstrate that differences in tube length arise primarily late in flower development, just prior to and following anthesis, and result from differences in both rate and duration of growth. Nectar tube growth rates increase over the course of development in both species, reaching a maximum just prior to anthesis. Flowers of P. odoratissimum reach anthesis 4–5 d earlier and attain a maximum tube growth rate of only 30 % of those of P. ionidiflorum. Unlike P. odoratissimum, nectar tube growth rates of P. ionidiflorum spike just before anthesis and, although transitory, this increased growth rate appears to contribute substantially to differences in final tube length between the species. Based on the data in Fig. 4C and D, we estimate that prior to the rapid increase in growth rate of P. ionidiflorum that occurs at 15 d, the tube growth rate is 1.1 mm d^–1 and the tube length is 15 mm. If the nectar tubes of P. ionidiflorum were to continue to grow with no change in rate until day 20, when they typically reach anthesis, then the final length would be approx. 20.5 mm, just two-thirds of the final observed length of approx. 30 mm (Table 1). Thus, the pre-anthesis spike in P. ionidiflorum growth rate, in addition to the longer duration of growth, contributes substantially to differences in final tube length between the two species.

Given that the contrasting tube lengths of P. ionidiflorum and P. odoratissimum result primarily from growth differences occurring in later stages of development, and that late organ growth is typically associated with cell expansion (Irish, 2008; Hepworth and Lenhard, 2014), cell expansion in the region of the receptacle surrounding the nectar tube should be greater for P. ionidiflorum than for P. odoratissimum. Cells of P. ionidiflorum are about 1.6 times as long as those of P. odoratissimum, indicating that cell expansion accounts for a significant portion of difference in length between these species. This difference in cell length, however, cannot entirely account for the 3-fold difference in tube length, and the species probably differ in the rate or duration of cell division as well. In contrast, for Aquilegia, the length of the freely projecting nectar spurs varies among species from 1 to nearly 16 mm (Puzey et al., 2012). In all species examined, cell division appeared to cease at the same spur length, and differences in final lengths were attributed solely to cell expansion. Similarly, in spurs of Centranthus (Mack and Davis, 2015) and Linaria (Box et al., 2011), cell divisions ceased early in development and cell expansion accounted for most of the elongation of the spurs. In the Polemoniaceae, cell size accounts for differences in corolla tube lengths of several species of Gilia, and Saltugilia, but the very largest flowers also had greater numbers of corolla tube cells (Landis et al., 2016). Clearly, more studies are needed to determine whether differences in cell division or expansion are more generally associated with evolutionary changes in nectar tube or spur lengths.

Although growth of P. odoratissimum flowers is truncated relative to P. ionidiflorum flowers, they are not scalar versions of each other, i.e. P. odoratissimum is not simply a smaller P. ionidiflorum. Allometric analyses show that although they are smaller at maturity, nectar tubes of P. odoratissimum flowers elongate faster relative to petal growth than those of P. ionidiflorum, but stop at a smaller size relative to the developing bud. This contrast in proportional growth indicates that differences between species do not arise from uniform patterns of cell division and expansion acting simultaneously across all floral organs but ceasing at different times. Rather, changes in tube development are somewhat independent of overall flower growth. More generally, such independence, or parcellation, of components of development may be critical for evolutionary diversification (Wagner, 1996; Klingenberg, 2008; Clune et al., 2013; Diggle, 2014; Strelin et al., 2016).

Homology of the nectar tube

The nectar tubes of Pelargonium have been referred to as spurs. In general, however, nectar spurs are hollow, slender, sac-like outgrowths of the perianth (Koopman and Ayers, 2005; Tucker and Hodges, 2005; Caris et al., 2006; Bernadello, 2007) that extend freely from their organ of origin and are not adnate to other floral whorls. Our developmental analyses are consistent with a receptacular origin of the nectar tube in Pelargonium; however, it is possible that the nectar tube evolved via fusion and subsequent displacement of an originally sepalar spur to a more interior position, as commonly inferred (Almouslem and Tilneybassett, 1989; Bakker et al., 2005; Bernadello, 2007; Rinbgelberg, 2012; Röschenbleck et al., 2014). This interpretation is unlikely given that other nectaries in the Geraniales are receptacular (Jeiter et al. 2017). Moreover, we find no evidence in the vascular patterns of these species to support an evolutionary origin of the nectar tube in Pelargonium as a spur adnate to the pedicel. Thus, our developmental and anatomical analyses support the earlier conclusions of Japp (1909) and Labbe (1964) that the nectar tubes of Pelargonium are deep single dorsal receptacular nectaries, not spurs or hypanthia.

CONCLUSIONS

The 3-fold differences in nectar tube length between two closely related species of Pelargonium are largely associated with changes in cell division and expansion that are manifest late in development. If the patterns we detected in two species are more general, then, by extension, the tremendous diversification of Pelargonium may be facilitated by the underlying lability of these late developmental processes. Flowers of different sizes, however, are not scalar versions, and growth dynamics of the nectar tube evolve independently of corolla size in P. odoratissimum and P. ionidiflorum. Finally, although the nectar tube of Pelargonium has been referred to as a spur or as part of a hypanthium, our data clearly demonstrate that tube development is fully integrated with the receptacle. This mode of nectar tube development is unique among plants.

SUPPLEMENTARY DATA

Supplementary data are available at https://academic.oup.com/aob and consist of Figure S1: epidermal peels of dorsal surfaces of the nectar tubes in P. ionidiorum and P. odoratissimum.