Flower longevity and size are coordinated with ecophysiological traits in a tropical montane ecosystem

The size of a flower and the duration it stays open and functional play central roles in the reproductive success of animal-pollinated plants. Larger and more conspicuous flowers tend to attract pollinators better, increasing the probability of outcrossing. Likewise, the time the corolla is on display for pollinators enhances pollen dissemination and the probability of outcrossing (Primack, 1985; Ashman & Schoen, 1994). Despite the importance of floral attractiveness to plant reproduction, the physiological strategies associated with flower size and longevity remain poorly defined (Lambrecht, 2013; Roddy et al., 2021). These strategies relate to how carbon, water, and nutrients are allocated to build and maintain the flower throughout its lifespan, which would encompass the resource costs of the mature physical structure, such as flower biomass (Roddy et al., 2023), the biochemical costs of mobilizing resources to build and expand the flower (Savage, 2019), the water, and carbon lost to transpiration and respiration (Teixido & Valladares, 2014), as well as the resources needed to produce pigments (Van Der Kooi et al., 2019a), nectar (Southwick, 1984), and volatile organic compounds (Adebesin et al., 2017). As a first-order approximation, the water and carbon costs involved in building and maintaining the flower could be substantial and be key components influencing floral physiological strategies. Flowers could be constructed identically per unit size, such that larger flowers would require, in total, more resources. Similarly, resources required to maintain a flower per unit of time could be constant, such that longer-lived flowers would require, in total, more resources. However, selection may act to minimize the overall physiological costs of floral production and maintenance, which could translate into altering the amount of resources required per unit of time or unit size. In this case, floral longevity and display size may be coupled to trait combinations that decrease resource expenditure while optimizing sexual reproduction (Ashman & Schoen, 1994). Across plant organs, longevity has been shown to scale with resource allocation patterns (Reich, 2014). In leaves, longevity is linked to the time required for carbon investments made during construction and the respiratory demands over the lifetime to be repaid by photosynthetic carbon assimilation (Reich et al., 2003; Wright et al., 2004). Long-lived leaves tend to exhibit high investments in structural carbon to cope better with physical damage and harsh conditions, which come at the cost of having low-instantaneous photosynthetic rates (Reich et al., 1992, 2003; Wright et al., 2004). Flowers, alternatively, are mostly heterotrophic and do not typically assimilate substantial amounts of carbon, releasing them from the need to maintain high rates of resource flux to support autotrophic metabolism (but see Brazel & Ó'Maoiléidigh, 2019, for examples of photosynthetic activity in reproductive organs). Rather, flower resource allocation must ensure reproduction while avoiding herbivory and resisting abiotic stress (Roddy et al., 2021). Because flowers are typically short-lived compared to leaves, selection may have favored making them cheap with a relatively low-biomass investment but a high-water mass investment for a given display area (Zhang et al., 2017; An et al., 2023; Roddy et al., 2023). However, to attract pollinators, flowers are typically displayed at the outer edge of the plant canopy, a placement that exposes them to the hottest, driest, sunniest, and windiest conditions experienced by the plant, all of which can elevate water loss and impair flower water status, potentially precluding successful pollination (Patiño & Grace, 2002; Bourbia et al., 2020; Carins-Murphy et al., 2023; Aun et al., 2024). In this sense, flowers should be mechanically sound, physiologically functional, attractive, and cheap to build, which imposes multiple selective demands on their construction and maintenance costs (Roddy et al., 2021). Due in part to their low-investment of carbon (e.g. low-petal dry mass per unit area (PMA) and petal thickness (PT)), flowers are limited in their capacity to restrict water loss across the cuticle, making them particularly vulnerable when water is limited (Zhang et al., 2018; Bourbia et al., 2020; Carins-Murphy et al., 2023; Roddy et al., 2023). Because most flowers have few or no stomata, the minimum or residual conductance (i.e. the minimum rate of water loss, gmin) is an important determinant of flower water balance (Roddy et al., 2016, 2023; Duursma et al., 2019; Aun et al., 2024). This minimum conductance also interacts with other floral traits, such as size and color, to influence flower temperature (Roddy, 2019). Due to low-vein density, most flower petals have a low capacity to transport liquid-phase water. Consequently, they may rely predominantly on stored water (i.e. hydraulic capacitance) to supply much of the water they lose (Roddy et al., 2013, 2016, 2019; Zhang et al., 2018). As a result, flowers have relatively high-water contents per unit area (Warea) and per unit of dry mass (Wmass) (Roddy et al., 2023). High-water content makes flowers cheap to build and biomechanically robust, and it can also delay desiccation. Warea and gmin can be combined to estimate the time required for full desiccation when there are no new water inputs. The average steady-state water residence time (τ) is defined as τ = W area g s D $$ \tau =\frac{W_{\mathrm{area}}}{g_{\mathrm{s}}D} $$ , where gs is the surface conductance and D is the vapor pressure deficit driving transpiration (Farquhar & Cernusak, 2005). For flowers with few or no stomata, the surface conductance is predominated by the residual conductance, gmin. Because flower performance is hindered if flowers desiccate, these physiological traits may be linked to the amounts of water and carbon allocated to build cellular structures. Here, we investigated whether biomass and petal water content (PMA, PT, Warea, and Wmass) and traits associated with water conservation (gmin and τ) are coordinated with flower longevity (FL) and petal surface area (PA). We measured traits of flowers from 19 plant species belonging to 15 families of monocots and eudicots native to the Brazilian campo rupestre ecosystem, a tropical montane biodiversity hotspot (Fig. 1; Supporting Information Table S1, Silveira et al., 2016; Miola et al., 2021). In this ecosystem, plant communities flower continuously throughout the year (Santos De Oliveira et al., 2021), and most species are pollinated by generalist bees (Table S1; Monteiro et al., 2021). Among these 19 species, flower longevity spanned from less than half a day for Cephalostemon riedelianus, Chamaecrista ramosa, Evolvulus lithospermoides, Ipomea procurrens, Pseudotrimezia juncifolia, and Xyris nubigena, up to 7 d for Kielmeyera regalis. Traits linked to petal biomass investment were associated with variation in longevity. Flower lifespan scaled positively with PMA (r2 = 0.39, P = 0.004; Fig. 2a) and petal thickness (r2 = 0.30, P = 0.01; Fig. 2b). Longer-lived flowers were, therefore, more expensive to build, supporting previous evidence (Zhang et al., 2017; Genty et al., 2023). Petal dry mass per area ranged widely, from 6.94 in E. lithospermoides to 95.86 g m−2 in Kielmeyera petiolaris. Similar to leaves, variation in PMA can be due to variation in petal thickness, meaning that thicker petals could have higher biomass and potentially higher water content for a given projected surface area (Witkowski & Lamont, 1991; Vendramini et al., 2002; Poorter et al., 2009). In our dataset for flowers, PMA was strongly related to petal thickness (r2 = 0.79, P < 0.001, Fig. 2c) and Warea (r2 = 0.69, P < 0.001, Fig. 2d). Thicker petals can hold more water per unit area, which could lengthen water residence times (r2 = 0.58, P < 0.001, but not significant when accounting for phylogeny, Table 1). These results suggest that long-lived flower tissues may be thicker and more durable – and, thus, more expensive to build – but they may better maintain biomechanical function and conserve resources throughout their lifespan (Zhang et al., 2017; Roddy et al., 2019; An et al., 2023). The lack of relationship between Warea and longevity despite its strong association with PMA (r2 = 0.69, P < 0.001, Table 1) and PT (r2 = 0.76, P < 0.001, Fig. 2e) reinforces the idea that increasing longevity is accomplished primarily by investments of carbon rather than water (Wright et al., 2004; Kikuzawa et al., 2013; Reich, 2014; Zhang et al., 2017). These patterns of dry matter were not necessarily reflected in water conservation, as quantified by gmin. A greater biomass investment per unit of floral display could confer better resistance to water loss if the biomass were allocated to structures that restrict water permeability, such as the cuticle (Duursma et al., 2019; Machado et al., 2021). However, variation in PMA and PT did not explain variation in gmin (r2 = −0.04, P = 0.38, and r2 = −0.07, P = 0.24, respectively) (Table 1), suggesting that floral biomass is unrelated to the rates of water loss. Instead, gmin was most strongly linked to PA (r2 = 0.60, P < 0.001; Fig. 2f), indicating that larger flowers better reduce water loss. Because gmin is normalized per unit evaporative surface area, its negative scaling with flower size reiterates that the difference between a large and a small flower is not just how many resources have been allocated to the entire flower but, even more importantly, that each unit of petal area is being built differently depending on whether it is in a large or a small flower. The negative scaling between PA and gmin resulted in larger flowers having longer water residence times (r2 = 0.67, P < 0.001; Fig. 2g). τ ranged from c. 15 min in the small flower of X. nubigena up to 24 h in K. regalis, and this variability was driven predominantly by gmin (r2 = 0.56, P < 0.001, Fig. 2h). Interestingly, τ was not related to longevity, as would be expected if most of the water transpired during a flower's lifespan came from stored water imported during floral expansion. Instead, even though flowers have low-vein densities and low-hydraulic conductance (Roddy et al., 2016), our results suggest that they transport enough water to replace their water content multiple times during their lifespans, consistent with data showing that most of the water influx to flowers occurs during anthesis (McMann et al., 2022). Nonetheless, longer τ indicates that physiological processes are more decoupled from dynamic environmental variation (Farquhar & Cernusak, 2005; Simonin et al., 2013; Roddy et al., 2018). For example, flower temperature depends on atmospheric conditions as well as a variety of floral traits, including PA and gmin (Roddy, 2019). Like leaves, larger flowers would have a thicker boundary layer, leading to lower convective heat transfer and hotter flower temperatures (Ackerly et al., 2002; Leigh et al., 2017; Wright et al., 2017). Though flowers often benefit from being warmer than the surrounding air because warmer flowers can better attract and reward pollinators (Sagae et al., 2008; Van Der Kooi et al., 2019b), flowers can become too hot, hastening desiccation or hampering pollen and seed development (Patiño & Grace, 2002; Roddy, 2019; Carins-Murphy et al., 2023). In addition to being hotter, larger flowers would also have more stable temperatures because of lower boundary layer conductance and higher thermal mass. The lower gmin and longer τ among larger flowers would further limit the cooling capacity of larger flowers, reinforcing the effects of flower size on flower temperature and thermal stability. Yet, larger flowers may be more vulnerable to overheating than smaller flowers. Thus, floral morphological and physiological traits interact in complex ways to influence pollinator attraction without excessive heating that would cause floral failure. Our results suggest that across species, flower biomass and water conservation traits are strongly linked to two traits important to pollination: flower longevity and size. The allocation of resources within flower tissues, whether towards building structures with a higher carbon content that last longer or increasing flower size at the expense of thermal regulation, indicates that increasing the probability of pollination of individual flowers may be related to trade-offs in floral physiological traits. As these trade-offs are likely to interact with abiotic factors, variation in abiotic conditions may affect floral attractiveness in campo rupestre species. For instance, hotter temperatures could disproportionately impact larger-flowered species due to their lower boundary layer and surface conductances. Alternatively, long-lived flowers may exhibit trait combinations that reduce their costs to offset higher maintenance respiration in hotter climates (Arroyo et al., 2013; Pacheco et al., 2016; Song et al., 2022). Moreover, as environmental conditions and resource availability vary seasonally, resource allocation to flowers might be associated with patterns of flowering synchronicity and intensity over time (Arroyo et al., 2013; Oleques et al., 2017; Nepal et al., 2024). However, studies formally addressing the effects of climate change on flower performance and phenology are still needed. In this regard, understanding how floral morphological and physiological traits influence floral performance and the likelihood of pollination is central to predicting plant responses to climate change. Our field sampling occurred in the campo rupestre ecosystem in the southern portion of the Espinhaço mountain range at Serra do Cipó, Minas Gerais, Brazil (19°16′56.8″S, 43°35′31.8″W, 1145 m asl). The campo rupestre is an open ecosystem composed of a mosaic of habitats embedded in a matrix of grasses and quartzitic/sandstone outcrops, occurring mostly on mountaintops of Brazil. It is recognized for its high diversity and endemism of species representing c. 15% of the angiosperm Brazilian species diversity (Silveira et al., 2016). The study area is considered a Cwb (subtropical highland climate) according to the Köppen classification, characterized by distinct, dry winters (May–September) and rainy summers (October–April) (Alvares et al., 2013). We sampled during the rainy season, from January to April 2023. We collected flowers only in the morning from 07:00 to 09:00 h to avoid sampling during the hottest periods of the day, which could drive floral desiccation. We selected newly opened flowers on sun-exposed branches that were visually healthy with no damage. We excised flowering shoots in the field, sealed them inside plastic bags, and transported them to the lab in a thermal box. Once in the lab, samples were placed inside a refrigerator until measurement. For each trait type (biomass and water conservation traits), we sampled at least eight flowers per species from at least four individuals. We dissected the petals, sepals, and reproductive structures and weighed each component separately to estimate fresh mass (precision of 0.0001 g, Sartorius BP 210 S). All structures were scanned on a flatbed scanner (resolution of 300 dpi). The projected surface area of each structure was measured using ImageJ software (Schneider et al., 2012; Roddy et al., 2021). For the trait calculations, we used the mass and area of the conspicuously colored perianth – that is, petals for the eudicot species and tepals for monocot species. We measured petal thickness with a digital thickness gauge (precision of 0.01 mm, Mitutoyo SERIES No. 700) and averaged three measurements over the petal surface (PT, mm; Pérez-Harguindeguy et al., 2016). Lastly, samples were dried at 70°C for 72 h and weighed to obtain dry mass for each structure. From these measurements, we calculated petal dry mass per unit area (PMA, petal dry mass divided by one-side petal projected surface area, g cm−2); petal water mass per unit area (Warea, petal fresh mass divided by one-side petal projected surface area, g cm−2); and petal water mass per unit dry mass (Wmass divided by PMA, g g−1) (Pérez-Harguindeguy et al., 2016; Roddy et al., 2021). To estimate residual conductance (gmin), we excised flowers at the peduncle base and sealed the cut surface with glue. We hung flowers on 100 g load cells interfaced to dataloggers (precision of 0.001 g; SparkFun OpenLog Artermis, SparkFun Electronics, Niwot, CO, USA) that measured sample mass every 2 s for c. 40 min. We placed samples inside a dark, custom-built thermal box with temperature and humidity sensors that were also connected to dataloggers and sampled every 2 s. The chamber had fans circulating air vigorously inside the box to minimize the boundary layer resistance around the samples. As samples slowly dissected, their mass decreased, from which we estimated the rate of water loss. After the 40-min sampling was complete, samples were removed from the box, dissected, and scanned for subsequent measurement of surface area. gmin (mmol m2 s−1) was calculated based on standard methods (https://prometheusprotocols.net/function/gas-exchange-and-chlorophyll-fluorescence/stomatal-and-non-stomatal-conductance-and-transpiration/minimum-epidermal-conductance-gmin-a-k-a-cuticular-conductance/) by calculating the rate of sample mass change over time, accounting for the vapor pressure deficit driving water loss, and then normalizing the conductance by the two-sided sample surface area. We calculated τ as Warea divided by gmin, assuming a constant vapor pressure deficit of 1 kPa (Simonin et al., 2013; Roddy et al., 2018, 2023). Higher VPD, such as those normally encountered during naturally varying conditions, would increase the rate of desiccation and shorten τ. To estimate the length of time a flower remains open and functional, we tagged at least 40 individual flower buds from at least four different plants of each species. We surveyed each flower twice daily, in the morning and at twilight. Longevity was quantified as the number of days from the flower's opening until the corolla wilted, changed color, or petals started falling (Primack, 1985; Roddy et al., 2021). At the species level, flower longevity was calculated by averaging the time each flower remained open. Six species had flowers that lasted less than half a day, that is they were just opening during the morning survey and were completely wilted by the twilight survey of the same day. Because our surveys were temporally coarser than the longevities of these six species, we tested how floral longevity variation impacted our results by using a randomization procedure. For each of these six species, we randomly generated 1000 values of longevity ranging from 0.2 to 0.4 d. For each iteration, we calculated the standard major axis regression (SMA) between longevity and each trait. In this analysis, all iterations produced SMA regressions with P-values < 0.05 (Fig. S1), suggesting that unmeasured variation in longevity among these short-lived flowers had no substantial effect on our conclusions. We performed all statistical analyses using R (R Core Team, 2021). We created a dated phylogeny using the v.phylomaker package (Jin & Qian, 2019) after checking species names and families using the taxize package (Chamberlain & Szöcs, 2013). To account for the effects of shared evolutionary history on trait relationships, we used phylogenetic independent contrasts implemented in the ape package with log10-transformed values (Paradis et al., 2004). To estimate the scaling relationships between traits, we used standard major axis regressions implemented in the smatr package (Warton et al., 2012) on log10-transformed trait values. We thank W. Justino and C. Rago for field assistance and G. W. Fernandes for facilitating field access. Financial support came from the Tinker Foundation, Judith Evans Parker Travel Award funding, the American Philosophical Society, and from US NSF grant no. CMMI-2029756 to ABR. None declared. DCP and ABR conceptualized the study and analyzed the data. DCP collected the data and wrote the first version of the manuscript. Both authors edited the manuscript. All data are available and provided in the Supporting Information. Fig. S1 Distribution of standard major axis regressions P-values. Table S1 Species-level flower traits. Please note: Wiley is not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.