Introduction:
Plant reproduction underpins global food production and is fundamental to plant genetic diversity and proliferation (Hedhly et al., 2009; Soltis & Soltis, 2014). As atmospheric temperatures rise under the effects of climate change, causing soils to dry faster, the water available to plants is becoming increasingly limited (Brodribb et al., 2016). Water limitation and its damaging effect on the water transport capacity of plants has been linked with mass plant mortality around the globe (Prasad et al., 2008; Choat et al., 2012; Anderegg et al., 2016; Lesk et al., 2016; Adams et al., 2017; Brodribb et al., 2020). Despite the serious consequences of reproductive failure to agriculture and conservation, the effects of water stress on plant reproduction are not well resolved (Roddy, 2019; Bourbia et al., 2020; Zhang et al., 2020). Hydraulic traits have been used to model plant mortality in drought (Sperry & Tyree, 1988; Choat et al., 2012; Vilagrosa et al., 2012; Urli et al., 2013; Anderegg et al., 2015; Sperry & Love, 2015; Choat et al., 2018), yet such traits in reproductive and floral organs are not well integrated into whole plant hydraulic vulnerability models.
Plant tissue damage and mortality from drought have been closely linked with the cavitation of xylem cells, caused by the invasion of air into vascular conduits under tension (negative pressure) (Sperry & Tyree, 1988; Tyree & Sperry, 1989). Decreases in soil water availability cause a drop in plant water potential, resulting in cavitation damage to the vascular system, progressively isolating plant tissues from their water source and ultimately resulting in tissue death (Savi et al., 2015; Brodribb et al., 2016). More cavitation-resistant xylem maintains water transport capacity longer into a period of drought than comparatively less resistant xylem (Sperry & Tyree, 1988; Tyree & Sperry, 1989). Relationships between xylem cavitation and plant damage have been widely characterized in woody tissues, but only recently have the implications of xylem cavitation been considered in the context of plant reproductive success (Zhang & Brodribb, 2017; Roddy et al., 2019; Bourbia et al., 2020).
Water is essential for plant reproduction, sustaining cell turgor and growth. In tomato, water accounts for more than 90% of ripe fruit mass, and fruit volumetric growth is primarily driven by water accumulation (Ho, 1980; Ho et al., 1987; Matthews & Shackel, 2005). Determining how water enters the fruit has key relevance for understanding reproduction in the context of whole plant water relations (Ho et al., 1987; Windt et al., 2009; Li et al., 2021). Water supplying the reproductive tissues must be delivered via xylem or phloem, although the relative contribution of these cell types to flowers and fruit has attracted debate. Some water would necessarily be delivered by the phloem in parallel with soluble sugars. If fruit water were supplied predominantly by phloem, reproductive tissues could theoretically be largely disconnected from apoplastic water stress (Chapotin et al., 2003; De La Barrera & Nobel, 2004). Alternatively, a majority xylem supply of water to fruit would expose reproductive tissues to the vascular tension experienced by the broader plant under drought stress (Zhang & Brodribb, 2017; Bourbia et al., 2020). If fruit rely on xylem water delivery, then major cavitation in the supplying xylem (in the peduncle and stem) would sever the supply of water for growth, theoretically damaging reproductive success. The relative contribution of xylem to the water balance of fruit and flowers has not been conclusively shown across species (Chapotin et al., 2003; De La Barrera & Nobel, 2004; Feild et al., 2009a; Feild et al., 2009b; Zhang & Brodribb, 2017; Zhang et al., 2018; Roddy et al., 2019). This inconsistency is particularly apparent in tomato, with xylem estimated to supply between 10% (Ho et al., 1987) and 90% (Windt et al., 2009) of fruit water content (van Die & Willemse, 1980; Hanssens et al., 2015). Recent magnetic resonance imaging (MRI) studies suggest a dominant xylem supply to tomato fruit, with 90% of fruit water found to be delivered via xylem (Windt et al., 2009), an observation supported across other species (van Die & Willemse, 1980; Windt et al., 2009; Hanssens et al., 2015). Determining the mode of water delivery is highly relevant to understanding fruit vulnerability to drought-damage, better placing plant reproduction in the context of whole plant water relations.  Here we use phloem girdling to establish the importance of xylem delivered water for fruit growth in water stressed plants, an essential first step before examining the physiology of the reproductive tissue xylem under drought.
Evidence suggests that reproductive tissues experience xylem cavitation in drought conditions  (Lambrecht, 2013; Roddy et al., 2018; Zhang et al., 2018; Bourbia et al., 2020). These observations highlight the need to include reproductive organs in hydraulic models to understand the broader patterns of xylem vulnerability throughout the plant (Bourbia et al., 2020). Based on the hydraulic vulnerability segmentation hypothesis, distal, short-lived organs requiring a low carbon investment should be more vulnerable to cavitation than non-redundant organs, helping to prolong plant survival in drought (Tyree & Ewers, 1991; Zimmermann, 2013; Johnson et al., 2016). By integrating reproductive tissues into whole plant hydraulic vulnerability studies, damage to reproductive organs in acute water stress may be compared with other organs based on their resistance to cavitation, thus illustrating their hydraulic priority during drought relative to other tissues (Bourbia et al., 2020; Zhang et al., 2020).
Recent data suggest that flowers are ranked low in the hierarchy of water supply priority during drought. This makes sense in the woody species examined where the high evaporative cost of flowering (Galen et al., 1999; Lambrecht, 2013) makes flowers an avoidable liability under water stress. It has been shown that in woody and perennial species flower xylem cavitates prior to other tissues under water stress, leading to flower shedding, while important vegetative structures such as stems and leaves are maintained (Tyree & Ewers, 1991; Bourbia et al., 2020; Zhang et al., 2020). This order of tissue preservation was evident from different organ xylem vulnerabilities to cavitation, with peduncles found to be highly vulnerable, effectively isolating leaky inflorescences during water stress (Bourbia et al., 2020; Zhang et al., 2020).  Applying the same adaptational logic to annual plants, it might be expected that the prioritization of reproductive tissues would be different given that these species only have a single opportunity to reproduce before death. We hypothesize that the reproductive tissues of annuals should be hydraulically prioritised under acute water stress through increased cavitation resistance in the xylem supply to fruit (via the peduncle and stem), relative to leaves (petiole and lamina). In this way fruits may benefit from the capacitive effects of water flowing from cavitated or collapsing tissue in leaves and stems (Higuchi & Sakuratani, 2005; Higuchi & Sakuratani, 2006; Hölttä et al., 2009), prolonging the time available for seed development during terminal drought. Provided a xylem connection between reproductive and vegetative tissues remains intact, water derived from the early cavitation of leaves could be accessible to reproductive tissues. This may be enabled by the capacity of fruits to exert low water potential through osmotic concentration of solutes, thus ensuring water flow towards the fruits even during acute water stress (Riboni et al., 2013; Riboni et al., 2016; Shavrukov et al., 2017).  
Here, we investigate the distribution of xylem vulnerability among reproductive and vegetative tissues in a herbaceous annual to determine the hydraulic priority of reproductive tissues relative to vegetative tissues under water-limiting conditions. Domesticated tomato was selected as a well-documented test species with an annual life history (although the ancestral species are likely to have been more perennial in growth habit) and large reproductive water requirements (Bussières, 1994; Bertin, 2003; Van Ieperen et al., 2003; Shameer et al., 2020; Li et al., 2021). We hypothesise that fruit growth will be hydraulically prioritized during drought through higher resistance to xylem cavitation in the stem and reproductive water supply (peduncle) compared with leaves (lamina and petiole); and that fruit growth may be maintained even during water stress by drawing on water provided by the capacitance and cavitation of less resistant tissues.

Materials and Methods:

Plant material and growing conditions
Tomato plants (Solanum lycopersicum var. Rheinlands Ruhm) were grown from seed in glasshouse facilities at the University of Tasmania, Australia. Plants experienced average day/night temperatures of 20 (±0.21SE)/15(±0.20SE) °C, an (automated) 18-hour photoperiod to ensure flowering, a relative humidity of approximately 40%, and were watered to field capacity daily to avoid drought stress. Seeds were germinated in punnets of fine pasteurised soil and then transferred to 2-l pots filled with potting mix (medium 7:4 mix of composted fine pine bark and coarse washed river sand). Plants were around 3 months old at commencement of experiments.
Plant preparation
Cavitation in the lamina, petioles, petals, and peduncles was monitored simultaneously in intact plants to determine the relative xylem vulnerability and order of cavitation among reproductive and vegetative organs within individual plants. Measurements of stem vulnerability were completed separately on replicate plants to avoid excessive manipulation (see below). Cavitation measurements were conducted on five replicate plants (3+ months old, ~1m tall) with fully expanded leaves and at least two inflorescences with separate peduncles. Well-watered plants were removed from pots and their roots were gently rinsed with water to remove most of the soil. This ensured complete cavitation of all tissues within 8 days. This duration ensured plant dehydration rates were fast enough to capture embolism formation but slow enough to maintain water potential equilibrium between slowly evaporating tissues (Bourbia et al., 2020).
Water potential monitoring
Once prepared, plants were transferred to the laboratory and allowed to dehydrate under artificial light (around 30 µmol quanta m-2 s-1) at a room temperature of 22oC. A stem psychrometer (ICT international, Armidale, NSW, Australia) was installed midway along the tomato stem above the first undamaged fully expanded leaf. A 20×30 mm section of stem cuticle and epidermis was gently removed by hand to expose the xylem. The xylem was then rinsed with distilled water and gently dried to remove tissue debris and potential solute contaminants. The stem psychrometer was clamped onto the stem parallel to the xylem. Parafilm (Bemis Co Inc, USA) was used to seal the psychrometer chamber to the stem and to cover any sections of stem with damaged cuticle. The psychrometer was set to log stem water potential (Ψs, MPa) every 20 min, with a Peltier cooling time of 10 seconds. The final cavitated water potential of tomato was not negative enough to require resetting of cooling time, with 10 seconds giving a stable reading of the wet-bulb temperature throughout drying. Ψs from the psychrometer was plotted over time and determined to decrease at a stable rate. A regression was fitted to these data and a linear relationship was determined; this relationship was used to compare cavitation resistance between organs. Ψs measurements made with the psychrometer were compared with measurements of leaf water potential assessed on detached leaves using a Scholander pressure chamber (PMS, Albany, OR, USA). Values agreed (±0.1MPa) until around 25% leaf cavitation, at which point the pressure chamber readings became unreliable due to leaf damage. After the point of stomatal closure, pressure gradients within the plant should collapse due to low transpiration rates. Given this assumption, the stem psychrometer Ψs was applied to the whole plant, with the pressure bomb values used as confirmation for the initial relationship (understanding that the assumption of water potential homogeneity breaks down as cavitation approaches 100%).
Organ xylem vulnerability to cavitation
The optical vulnerability method (OV) was used to quantify xylem vulnerability (Brodribb et al., 2015). This technique is minimally invasive and allows simultaneous monitoring of xylem cavitation throughout the plant. Data produced using the OV technique has been shown to correspond very closely with vulnerability data collected using hydraulic methods and x-ray computed tomography in tomato (Skelton et al., 2017) and several other species (Gauthey et al., 2020). A combination of custom-made “cavicams” (opensourceov.org) and a light microscope were used simultaneously on each plant to monitor cavitation during dehydration.
Leaf lamina
Cavicams were secured to one fully expanded lamina per plant to view an area approximately 5×5 mm that included secondary and tertiary veins. Leaves were illuminated from underneath by LEDs (white-light emitting diodes). Images were acquired every 2 min for 8 days. Image focus was adjusted as necessary.
Petiole
Cavicams were installed on the petiole of one adjacent leaf per plant close to the main stem. A window approximately 15×8 mm was made by carefully removing the epidermis and cuticle on one side of the petiole with a sharp razor blade to expose the xylem. Transverse sections of additional petioles, peduncles and stems were examined prior to testing to determine xylem depth from the surface. The exposed xylem was coated in hydrogel (Tensive Gel; Parker Laboratories Inc., Fairfield, NJ, USA) to improve light transmission and reduce evaporation. Petioles were then secured in cavicams and images were acquired as described above using reflected light.
Peduncle
Cavicams were installed on one inflorescence per plant with flowers and buds but no fruit. A window approximately 15×5 mm was made by carefully scraping away the epidermis and cuticle on one side of the peduncle with a fingernail to expose the xylem. A different method was used for peduncles because razor blades easily cut too deep into the vascular tissue. The exposed xylem was coated in hydrogel and then secured in a cavicam which acquired images using reflected light.
Petal
A Leica DFC450 digital camera, mounted on a Leica M205A stereomicroscope (Leica Microsystems, Wetzlar, Germany) was used to capture images of one petal per plant from an intact apical inflorescence. The petals of a newly opened flower were taped to a glass plate to prevent movement and deformation during dehydration and image capture. Light was transmitted through the petal illuminating the midrib xylem bundle, and images were captured as described above.
Stem
Four additional plants of the same age and size were used to measure stem vulnerability to cavitation. Stems were measured on separate plants so that cavicams could be positioned as far away from the psychrometer as possible to avoid damaging the xylem in the imaged field of view. This also avoided potentially damaging the xylem imaged by cavicams on nearby organs. Cavicams were installed on one section of stem per plant between the root collar and the first reproductive structures. A window approximately 20×10 mm was made by carefully removing the epidermis and cuticle with a sharp razor blade to expose the xylem. Water potential was monitored with a stem psychrometer as above and image capture protocols were as described for the petiole and peduncle.
 
Image analysis
Image sequences were analysed using IMAGEJ (NIH), see http://www.opensourceov.org/ for details. Briefly, the pixel value of each image were subtracted from the next in the sequence to quantify any differences between images. Pixel differences indicate a change in light transmission or reflection as water-filled vessels fill with air during cavitation events (Brodribb et al., 2016). Due to the frequency of image capture, cavitation events can be separated from image noise caused by movement such as leaf shrinkage, or drying of hydrogel, by removing differences that are not rapid pixel changes confined to the xylem. Noise was removed using pixel thresholding 30-∞ leaving only pixels associated with cavitation events. 30 pixels was smaller than the cavitation signature of the smallest vessel size, allowing exclusion of objects that were caused by noise or small movements. The total embolism area per image was calculated as the summed embolized pixels and expressed as cumulative embolism, a percentage of the total cavitated area in pixels per sequence. Area cavitated (% of maximum cavitated pixel area) was plotted against water potential to create vulnerability curves (Brodribb et al., 2017; Bourbia et al., 2020) and derive the water potential at which 50% of total xylem area was cavitated (P50). The P12 and P88 (water potential at which 12 and 88% total xylem area was cavitated) were also calculated for stems and peduncles. 
Tissue shrinkage/growth analyses
We monitored the shrinkage dynamics of petioles, peduncles, and fruit simultaneously within individual plants to investigate the distribution of water between vegetative and reproductive tissues during plant dehydration. The water potential of five additional plants was monitored during the same dehydration treatment as described above. Intact proximal regions of petioles and peduncles were secured in cavicams, requiring no cuticle preparation. Images were acquired using transmitted light to produce a silhouette every 10 minutes until the plant was fully desiccated (approximately 8 days). Cavicams were refocussed every day as the tissue dried. Images of intact tomato fruits approximately 1 cm diameter (older than 7 days, around 10% of their mature adult size) were captured using a Leica DFC450 digital camera. Single fruits were secured to the stage of a Leica M205A stereomicroscope (Leica Microsystems, Wetzlar, Germany) or a custom-built gantry holding a Nikon lens attached to a Raspberry PI camera and images were taken at 10 minute intervals until plants were fully desiccated. 
Images were processed in IMAGEJ (NIH). A threshold was applied to image sequences to isolate plant tissue from the background. As diameter was measured manually in IMAGEJ, a subset of images was made (every 50 images) to capture change in diameter over time (expressed as a percentage of the original diameter) and this was plotted against water potential. To ensure that fruit growth was dependent upon an intact vascular supply (i.e. that fruit did not expand after the vascular supply was severed using stored resources), the growth of a tomato fruit was also assessed before and after excision from the plant. A tomato fruit ~1 cm in diameter was monitored as above. After demonstrating initial positive expansion, the fruit was excised from the plant at the pedicel below the abscission zone using a razor blade, at a water potential of -1MPa. Images were captured and analysed using the same method described for intact fruits.
 Phloem girdling
We compared the growth rate of fruit with intact xylem and phloem to that of fruit with intact xylem and damaged phloem to determine the relative importance of xylem and phloem water delivery to fruit expansion during water stress. Phloem tissue was selectively killed (while preserving xylem function) by heat girdling the pedicel, following the method of Van De Wal et al. (2017) which was explicitly shown to sever the phloem connection while maintaining xylem delivered water flow in tomato. In brief, a length of insulated nichrome wire (length, 0.5 m; diameter, 0.25 mm) was coiled around the pedicel between the peduncle joint and the abscission zone, covering half of the total 8mm pedicel length. A DC electrical current (~7.5V) was applied via a transformer to raise the temperature to 75oC for 1 min. Temperature was measured using a thermocouple in contact with the plant tissue under the heating coil (Van De Wal et al., 2017). Measurements were performed on three plants, using paired fruit supported by the same peduncle, with one fruit per pair heat girdled and the other maintained as an ungirdled control. Following girdling, images were captured to monitor the change in diameter of both fruits as the plant dried. Plants were kept under laboratory light conditions (around 30µmol quanta m-2 s-1), and water potential was recorded using a psychrometer, attached to the stem as described above. Tomato fruits undergo a phase of cell division for 1-2 weeks post fertilisation, and then a period of cell expansion for 2 to 6 weeks, driven by the filling of cells with osmolytes and water (Ho et al., 1982; Gillaspy et al., 1993; Bertin, 2003; Renaudin et al., 2017; Shameer et al., 2020). Larger fruits (greater than 2 cm in diameter) were selected here to ensure cell expansion was responsible for growth rather than cell division.
Petiole and peduncle xylem anatomy
Anatomical properties of petiole and peduncle xylem were compared to provide insight into any functional differences in water supply between tissues. The xylem lumen diameter of three petioles and three peduncles from each of three additional plants was measured to examine the association between organ xylem cavitation vulnerability and vessel size. Peduncle and petiole tissue was removed close to the organ base where they joined the main stem. 30 µm transverse sections were made using a sliding microtome (Leica Microsystems, North Ryde, NSW, Australia) from samples secured in position using a BFS-3MP freezing stage (Physitemp Instruments, Clifton, NJ, USA) and encased in a dilute sugar solution. Sections were stained with toluidine blue (0.5%) and mounted on glass slides with phenol glycerine jelly. Slides were photographed at 20× magnification using a Nikon Digital Sight DS-Fi2 camera (Tokyo, Japan) mounted on a compound microscope (DM1000; Leica Microsystems, Wetzlar, Germany). Vessel diameter was measured using IMAGEJ (NIH). Four representative fields of view of entire xylem bundles were captured for each transverse section. A binary image was made of each section, and vessel lumens were selected using the threshold function. This captured only lignified cells with a clear lumen. An ellipse was automatically fitted to each xylem vessel lumen using the ‘analyse particles’ function (fig. S1). A vessel diameter was calculated from each ellipse using an average of the major and minor axes. Vessels were ranked numerically by diameter and a minimum size threshold of 5μm was imposed to exclude fibres and axial parenchyma. After calculating the maximum relative conductance (