Flowers are leakier than leaves but cheaper to build

Abstract

Flowers are critical to reproduction in angiosperms and have been credited with promoting diversification and the rapid spread of flowering plants globally (Sanderson & Donoghue, 1994; Crepet & Niklas, 2009; Leslie et al., 2021). Although they are typically short-lived, flowers require resources, such as carbon, water, and nutrients, for their production and maintenance (Reekie & Bazzaz, 1987a,b; Ashman & Schoen, 1994; Song et al., 2022). Water, in particular, is used throughout development and anthesis for a variety of functions, including driving growth and expansion, keeping flowers turgid and on display for pollinators, providing rewards such as nectar, and for regulating temperature (Bazzaz et al., 1987; Galen et al., 1999; Patiño & Grace, 2002; Chapotin et al., 2003; De la Barrera & Nobel, 2004; Roddy & Dawson, 2012; Roddy, 2019; Treado et al., 2022). Additionally, flowers regularly lose water to the atmosphere, and this water loss may increase during hot and dry conditions often associated with droughts (Hew et al., 1980; Feild et al., 2009; Teixido & Valladares, 2014; Sinha et al., 2022). Flower water balance is, therefore, critical to flower function, yet surprisingly little is known about the mechanisms of water balance in flowers, how physiological traits related to water and carbon influence the costs of floral display, and how floral hydraulic traits affect drought responses (Roddy et al., 2016, 2021; Bourbia et al., 2020; McMann et al., 2022). The rate of water loss from flowers – and, indeed, from all aerial organs of plants – is ultimately determined by the atmospheric conditions that drive the net loss of water from the plant to the atmosphere (e.g. solar radiation, temperature, humidity, and windspeed) and by the structure of the epidermis, which controls the total surface conductance to water vapor (gt). Stomata in the epidermis are the primary pathway for water movement from plants to the atmosphere, and their sizes and densities influence maximum rates of transpirational water loss (Hetherington & Woodward, 2003; Franks & Beerling, 2009). Compared with leaves, flowers often have relatively few, if any, stomata on their petals and petaloid structures (Hew et al., 1980; Lipayeva, 1989; Roddy et al., 2016; Zhang et al., 2018). Under well-watered conditions, the high densities of stomata on angiosperm leaves allow transpiration rates from leaves to exceed those of flowers (Feild et al., 2009; Roddy et al., 2018). However, under drought conditions, leaf stomata close to limit water loss, causing any remaining water vapor flux to be due to the minimum epidermal surface conductance (gmin), which is due to the conductance of the cuticle and any incompletely closed stomata (Kerstiens, 1996; Duursma et al., 2019; Márquez et al., 2022). After drought-induced stomatal closure in leaves, water loss from flowers can be as high as or even exceed water loss from leaves (Sinha et al., 2022; An et al., 2023), suggesting that corolla gmin may hinder the ability of plants to maintain floral display during periods of water stress (Lambrecht, 2013; Buschhaus et al., 2015; Bourbia et al., 2020). Yet, despite the influence of gmin on flower and whole-plant hydration and its role in regulating flower temperature, gmin has been measured on flowers of only a few species (Patiño & Grace, 2002; Roddy et al., 2016; Roddy, 2019; Bourbia et al., 2020). Here, we compared flowers and leaves in a set of physiological traits that influence water balance, particularly during drought (Brodribb et al., 2007; Boyce et al., 2009; Simonin et al., 2013; Roddy et al., 2016, 2018; Duursma et al., 2019; Bourbia et al., 2020; An et al., 2023). We measured gmin, water content per unit projected surface area (Warea) and dry mass (Wmass), vein density (Dv), and dry mass per area (leaf mass per area, LMA, or petal mass per area, PMA) of flowers and leaves for over 100 species from 41 angiosperm families growing in a common garden to determine how these physiological traits differ among organs and influence the costs of floral construction and maintenance. Flower petals differed significantly in most of the water balance traits we evaluated (Fig. 1; Table 1). Water content per unit dry mass (Wmass), which is positively related to hydraulic capacitance (Ogburn & Edwards, 2012; Roddy et al., 2019), was significantly higher in petals than in leaves. Flower petals also had higher Warea than leaves, though this difference was not significant after accounting for shared evolutionary history, and the range of Warea among species was larger for petals than it was for leaves. With a mean of 12.68 mmol m−2 s−1, flowers had significantly higher gmin than their neighboring leaves. Leaf gmin had a mean of 4.65 mmol m−2 s−1, which was similar to the interspecific mean of a recent literature survey (mean of 4.9 mmol m−2 s−1; Duursma et al., 2019) and similar to some tropical leaves (Slot et al., 2021). Of the 101 species for which we measured gmin data for both leaves and flowers, only 27 species had flowers with lower gmin than leaves. The relatively high gmin of flowers highlights that flowers may contribute significantly to whole-plant water budgets during flowering periods (Lambrecht & Dawson, 2007; Lambrecht, 2013). In many species, flowers are positioned distal to leaves, resulting in leaf shading and suppressing foliar transpiration (Shen et al., 2009; Sonnentag et al., 2011). Additionally, because flowers are often positioned in the hottest, driest parts of the plant crown, their high gmin may translate into high rates of water loss (Roddy & Dawson, 2012). For example, previous work on avocado has shown that a combination of high flower transpiration and high total flower surface area resulted in flowers accounting for c. 13% of total canopy water loss (Whiley et al., 1988). Given that corolla gmin is, on average, higher than leaf gmin (Fig. 1), flower water loss can be as high as or even exceed water loss from leaves and potentially dominate total canopy transpiration (Lambrecht, 2013). More work is needed at the whole-plant scale to characterize whether and when flower water loss can detrimentally impact whole-plant water status and potentially precipitate hydraulic failure in vegetative organs (Nobel, 1977; Galen et al., 1999; Lambrecht & Dawson, 2007; Zhang & Brodribb, 2017; Bourbia et al., 2020). Using our measurements of gmin and Warea, we calculated water residence times ( τ $$ \tau $$ ) for leaves and flowers, assuming a constant vapor pressure deficit (VPD) of 1 kPa. Despite having slightly higher Warea, flowers had significantly shorter τ $$ \tau $$ than leaves, driven by their higher gmin (Fig. 1; Table 1). A shorter τ $$ \tau $$ among flowers suggests that when water loss exceeds water supply, for example, during drought, flowers would desiccate more rapidly than leaves due to their higher gmin. Under higher VPD than the mild 1 kPa we used, τ $$ \tau $$ would be even shorter. For example, increasing VPD to 2.5 kPa reduces the average τ $$ \tau $$ for flowers from 21.86 to 8.74 h and for leaves from 47.83 to 19.13 h. Under this scenario, the average flower would desiccate within a day without new water supply, consistent with previous reports for Calycanthus flowers during a heatwave (Roddy et al., 2018). Despite having higher water contents and hydraulic capacitance (Fig. 1; Roddy et al., 2019), high gmin may require that flowers have constant supplies of water to remain turgid. That gmin has been shown to scale with whole-flower hydraulic conductance reiterates the role of gmin in regulating flower water balance (Roddy et al., 2016). Despite early suggestions that flowers may be hydrated primarily by the phloem (Trolinder et al., 1993; Chapotin et al., 2003), the phloem may not contribute meaningful amounts of water to the overall flower water budget (Feild et al., 2009; Roddy et al., 2018; McMann et al., 2022), and given a relatively shorter intrinsic τ $$ \tau $$ flowers may need to remain connected to the xylem hydraulic system to avoid desiccation (Feild et al., 2009; Roddy et al., 2018). In leaves and flowers, liquid water is delivered primarily by the network of veins that traverse the leaf and petal. However, flower petals had significantly lower Dv than leaves (Fig. 1; Table 1; Roddy et al., 2013; Zhang et al., 2018), suggesting that flowers have a lower hydraulic conductance than leaves (Brodribb et al., 2007; Roddy et al., 2016). The lower Dv despite higher gmin reiterates that flowers may be particularly vulnerable to drought, when water loss may outpace hydraulic supply, leading to rapid declines in flower water potential that could cause declines in stem water potential (Bourbia et al., 2020). However, the higher Warea and higher hydraulic capacitance of flowers (i.e. high Wmass) would minimize changes in water potential despite their having a low hydraulic conductance. The high hydraulic capacitance of flowers would, therefore, suppress diurnal variation in corolla water potential, reducing the impact of excessive floral water loss on the water potentials of stems and leaves and reducing the likelihood that flower water potential would decline enough to cause embolism spread (Zhang & Brodribb, 2017; Roddy et al., 2018, 2019). Understanding how gmin, Warea, and Wmass of flowers interact with hydraulic traits of leaves and stems, and by extension whole-plant water and carbon balance, will be critical to better characterizing plant responses to changes in water availability. Although they had higher water contents than leaves, petals had significantly lower dry mass per area than leaves (Fig. 1; Table 1), which has implications for the biomechanics of flower petals. For short-lived structures like flower petals, reducing the costs of floral display has likely been favored by selection (Roddy et al., 2016; Olson & Pittermann, 2019; Roddy, 2019). Longer-lived flowers may incur higher carbon costs because they may need to withstand attack by floral enemies (Ashman, 1994; Roddy et al., 2021; Boaventura et al., 2022; Song et al., 2022). How biomass costs of flowers are related to other resources can be variable and context-dependent (Bazzaz et al., 1987; Reekie & Bazzaz, 1987b; Roddy et al., 2021), but the water and carbon costs may be coupled in important ways. Supplying more water to flowers would require a denser network of veins or larger diameter xylem elements, which are carbon-rich and potentially costly to produce. Similarly, better-limiting water loss by building thicker or denser cuticles could also require more carbon investment (Buschhaus et al., 2015; Cheng et al., 2019). Yet, because every molecule of carbon requires at least 400 molecules of water to be transpired (Nobel et al., 2005), for short-lived structures such as flowers, water may be relatively cheaper than carbon, suggesting that flowers may employ a hydrostatic skeleton rather than an expensive, carbon-based skeleton for structural support. We tested whether floral display is cheaper in terms of carbon due to higher initial investment of water by examining the relationship between water content and dry mass investment. Warea scaled positively with LMA (slope = 0.96 (0.84, 1.09), R2 = 0.56, P < 0.0001) and PMA (slope = 1.01 (0.88, 1.16), R2 = 0.52, P < 0.0001), with statistically indistinguishable slopes between organs (P = 0.47). However, flowers had a significantly higher intercept to this scaling relationship (t = 2.69, df = 98, P < 0.01). A higher intercept among flowers could result from a reduction in dry mass per area at constant Warea, that is, a leftward shift from the leaf regression in Fig. 2(a). Similarly, Wmass scaled negatively with both PMA (slope = −0.75 (−0.91, −0.63), R2 = 0.14, P < 0.001) and LMA (slope = −0.69 (−0.83, −0.58), R2 = 0.16, P < 0.0001) with slopes indistinguishable between organs (P = 0.36) but a significantly higher intercept among flowers (t = 2.57, df = 98, P < 0.05). The higher intercepts among flowers in these scaling relationships support the hypothesis that flowers rely on a hydrostatic skeleton maintained by high water content and cheap, thin cell walls that allow for relatively low dry mass per unit area and lower cell wall elasticity (Roddy et al., 2019). The combination of high water content and low dry mass would also explain why flowers have higher hydraulic capacitance than leaves, as thin cell walls with a low modulus of elasticity would allow large changes in cell volume with relatively small changes in water potential (Roddy et al., 2019). While there are numerous implications of higher gmin in flowers, it is important to consider why gmin is higher in flowers. We propose two alternative explanations for this pattern. First, if reducing gmin requires greater carbon investment (e.g. through additional cuticular waxes; Cheng et al., 2019), then the additional carbon required to further reduce water loss may not be worth paying in such a short-lived organ. Second, cuticle structure and composition may experience divergent selection for multiple functions. For the majority of angiosperm species, petaloid organs attract pollinators through visual cues (van der Kooi et al., 2019), humidity gradients due to water loss (von Arx et al., 2012; Dahake et al., 2022), or the release of volatile organic compounds (Dudareva et al., 2013). Corolla cuticles are structurally different from those of neighboring leaves (Jetter et al., 2008; Cheng et al., 2019), and cuticle structure can influence optical properties, volatile emission, and conductance to water vapor (Goodwin et al., 2003; Whitney et al., 2009; Buschhaus et al., 2015; Liao et al., 2021). Thus, pollinator selection on any of these functions may be tightly coupled to the mechanisms of flower water balance. Because climate change is affecting both aridity and pollinator abundances, the relative costs of producing and maintaining flowers may also change, suggesting that climate change may alter the selection dynamics on floral hydraulic traits (Thomann et al., 2013; Gallagher & Campbell, 2017; Kuppler & Kotowska, 2021). Understanding how shifting selective regimes may impact floral function and evolution will be important in understanding the future viability of flowering plants globally. Plants were sampled in 2011 and 2012 at the University of California Botanical Garden in Berkeley, CA, USA, where plants are maintained well-watered throughout the year. The species samples represent a phylogenetically and ecologically diverse set of species that vary in stature, habitat, and flowering phenology (Supporting Information Table S1). We sampled in the morning, between 08:00 and 09:00 h, by excising a flowering shoot and immediately recutting the cut end under water a few inches apical to the initial cut or, for large woody species, one node more apical from the initial cut. If leaves were not present on this shoot, we excised a leafy shoot in a similar way. Cut ends remained in water, and the shoots were kept in a bucket to shield them from desiccation during transport. Shoots were transported back to the laboratory within 1 h and kept in the dark before trait sampling. We measured leaves and flowers on the same plant, and for most species, we sampled one individual plant per species. While this sampling approach may not be used to characterize average trait values for each species, it is particularly well-suited to test whether organs differ in traits. Minimum surface conductance (gmin) was measured by excising a leaf at the petiole basis or a flower petal. When flower petals were fused, we excised all fused petals, keeping them intact to minimize cut surfaces. Immediately upon excision, the cut surfaces of the petals or leaf petiole were sealed with either cyanoacrylate glue or petroleum jelly. Samples were either hung or placed on a mesh screen in a dark cabinet with a fan blowing directly on them in order to maximize the boundary layer conductance. Every 5–20 min, samples were weighed on an analytical balance (Sartorius CPA225D, resolution = 0.0001 g) and the temperature and humidity inside the chamber recorded. After c. 10 measurements, each sample was either scanned or photographed for subsequent measurement of its area and then dried at 70°C for at least 72 h for subsequent dry mass measurement. For flower petals, scanning sometimes meant we had to dissect and flatten the petals and ensuring no petals were overlapping. Areas of scanned leaves and petals were measured using ImageJ by manually tracing the outlines of the sampled organs (Schneider et al., 2012). gmin was calculated from measurements of mass, temperature, and humidity using the spreadsheet available at https://prometheusprotocols.net/function/gas-exchange-and-chlorophyll-fluorescence/stomatal-and-nonstomatal-conductance-and-transpiration/minimum-epidermal-conductance-gmin-a-k-a-cuticular-conductance/. For samples that displayed nonlinear change in mass over time, we discarded the initial one to two measurements under the assumption that these were most likely to be artificially high due to incomplete stomatal closure. gmin was calculated from the remaining measurements by fitting a linear regression to the relationship between mass change and the atmospheric vapor pressure deficit. In most cases, one flower and leaf per plant was sampled, and previous measurements suggest that intraspecific variation in flower gmin is relatively small (Roddy et al., 2016). Leaf (LMA) or petal mass per area (PMA) were quantified from the area and dry mass measurements made on samples measured for gmin. Water content per area (Warea) was calculated as the difference between the initial mass and dry mass divided by the surface area of each sample, and the water content per dry mass (Wmass) was calculated as W area LMA $$ \frac{W_{\mathrm{area}}}{\mathrm{LMA}} $$ , converted to units of grams of water per gram of dry mass. Water residence time ( τ $$ \tau $$ ) was calculated as τ = W area E $$ \tau =\frac{W_{\mathrm{area}}}{E} $$ where E is the transpiration rate, which was calculated from gmin assuming a vapor pressure deficit of 1 kPa (Simonin et al., 2013; Roddy et al., 2018). Using a higher vapor pressure deficit reduces τ $$ \tau $$ . Most of the vein density (Dv) data have been published previously (Roddy et al., 2013) and sampling methods are briefly summarized here. For leaves, c. 1-cm2 sections from midway between the leaf midrib and margin, midway between the base and tip of the leaf were excised and immediately placed into 4% NaOH. To account for the high variability in vein density within a petal, we collected multiple 1-cm2 sections from throughout the petals and placed them in 4% NaOH. After 2–4 wk, leaves were washed in distilled H2O, transferred to a 3% bleach solution for c. 20 min, washed again in distilled H2O, and then placed in 95% ethanol. After clearing in NaOH, petals were washed in distilled H2O and transferred into ethanol, skipping the bleaching step. Once in ethanol, samples were briefly stained with Safranin O and imaged at ×5 to ×20 magnification under a compound microscope outfitted with a digital camera. One or two images per section from each of five to 12 sections per species were captured. For each image, the total length of veins was measured manually using ImageJ (v.1.44o; Schneider et al., 2012) and divided by the total area of the image to calculate vein length per area. The mean for each structure of each species was calculated and used for subsequent analyses. All analyses were conducted in R using We determined by examining and for each trait and species with data for any which were a total of species all traits. We to data and species rather than between organs in each trait were determined using To account for the of sampling species with shared evolutionary history, we used phylogenetically analyses. A was using & 2019) after higher using the R & 2013). were conducted using the function in the R 2012). and their and were conducted using the R et al., 2012). work was by from the of the University of Berkeley, a by to and on We at the University of California Garden for to plant and the and collected the and the the All the All data are available and in the Table trait values for leaves and flowers. is not for the content or of any Information by the than be to the The is not for the content or of any by the than be to the for the

Topics & Concepts