Members of the grass family Poaceae have adapted to a wide range of habitats and disturbance regimes across the planet. The cellular structure and arrangements of leaves can help explain how plants survive in different climates, but these traits are rarely measured in grasses. Further, most studies are focused on individual species or distantly related species within Poaceae.  While this focus can reveal broad adaptations, it also likely to overlook subtle adaptations within more closely-related groups (subfamilies, tribes). This study therefore investigated the scaling relationships between leaf size, vein density, and vessel size in five genera within the subfamily Pooideae. The relationship between leaf area and major vein number was consistent with previous findings (p < 0.05, slope = 0.72 +/- 0.24), as was the scaling coefficient of VLA (slope= -0.46 +/- 0.21). However, several genera exhibited novel anatomical relationships. In Poa and Elymus, minor vein number and leaf length were uncorrelated, whereas in Festuca these traits were positively correlated (slope = 0.82 +/- 0.8). These findings suggest there is important broad-scale and fine-scale variation in leaf hydraulic traits among grasses. Thus, future studies should consider both narrow and broad phylogenetic gradients.

Plant Material Selection and Germination

Table 1- list of the species germinated for data collection.

Five species were selected from each of five genera: Poa, Hesperostipa, Elymus, Festuca, and Bromus. The species selected in this study are all phylogenetically classified as part of the Pooideae subfamily; this was done to ensure that all species were separated by relatively recent evolutionary divergences. Species were also selected such that their habitats spanned a wide range of temperature and precipitation across North America. Plants in this study were grown from seeds provided by the USDA Germplasm Resource Information Network. Unfortunately, not all species germinated regardless of any pre-treatments we attempted and so we were not able to measure 5 species in all genera. Specimens used for gas exchange measurements were grown during the summer of 2016 in 35-cm-length Deep-pots (D60 series, Stuewe and Sons, Inc, Tangent, USA)  while the remaining specimens were grown in 10-cm-length “cone”-tainers (Ray Leach Cells 3, Stuewe and Sons, Inc, Tangent, USA), with a 70% to 30% mix of potting soil (Promix HP, Quakertown, USA) and fritted clay (Greens’ Grade Porous Ceramic Topdressing, Buffalo Grove, USA), respectively. This substrate was fertilized with slow-release fertilizer (Osmocote Plus, Scotts Miracle-Gro Company, Marysville, USA) at a ratio of 10 mL fertilizer L^-1 soil substrate. The soil substrate was saturated with water before seeds were planted, and specimens were misted until the full expansion of their 4^th leaf, then watered 3 times per week.

Anatomical Analysis

Whole-leaf samples (base to tip of lamina) were collected from five randomly selected individuals of each species for the purpose of anatomical examination. The third or fourth full leaf was harvested from each plant after full expansion. Samples were stored in a solution of formalin, acetic acid, ethanol, and deionized water (FAA fixative solution). Anatomical samples were collected halfway along the lamina length of each specimen, using a razor to cut a cross-sectional sample ~0.5 mm in thickness. The distance from the cross-section and the leaf tip were also recorded. Each cross-sectional sample was stained using safranin-o and fast green. Microscopic images of vein anatomy were taken of half the total lamina width using a ZEISS Axio Scope.A1 in conjunction with ZEN microscope software (Carl Zeiss Microscopy, Germany). Images were measured using Fiji open-source image analysis software. The number of vein orders of each species was quantified, and 1° and 2° veins were classified as ‘major’ veins and 3° and 4° (if present) were classified as ‘minor’ veins. We defined major veins as having at least two of the three following characteristics: 1) vein was at least 50% larger than the smallest vein, 2) vein had bundle sheath extension, 3) vein had at least two xylary vessel elements (i.e metaxylem lacuna) 100% larger than remaining xylary vessel elements. Although this criteria differs slightly from what others have used to define vein orders (Baird et al. 2021), it was developed to help us to objectively assign vein orders since leaf veins don’t always fall neatly into the orders defined previously. However, we are confident that what we defined as ‘major’ is consistent with previous research. The anatomical traits of up to five major veins and minor veins per leaf were measured, but all veins were counted and classified to a vein order. Vein length density was calculated as total vein length per unit area (cm cm^-2) by measuring the width of the leaf and then counting the number of veins in the leaf, then multiplying by the length of the leaf. The diameter of the vein including the bundle sheath (μm) was measured as the distance from the outer edge of the bundle sheath to the opposite outer edge. Vein diameter (D_vein, μm) was measured from the innermost edge of the bundle sheath to the opposite inner edge. Diameter of the bundle sheath cells (D_BS) was then estimated by subtracting vein diameter measured from the innermost edge from the vein diameter measured from the outermost edge of the bundle sheath. The diameters of vessel elements (D_vessel, μm) were measured as the distance between the inner edge of the vessel element cell wall to the opposite inner edge. Vessel element wall thickness (WT; μm) was also measured. ‘MAJ’ is added to each subscript when the data reported was measured on major veins and ‘min’ added to subscript when the data represents the minor veins.

Climate Envelope Analysis

Species distribution data were collected from the Global Biodiversity Information Facility (GBIF). Climate data were obtained from weather stations closest to the location where seed was harvested as reported by GBIF. In addition to weather station observations, gridded and interpolated climate data were retrieved from WorldClim at 0.1 degree resolution. For each reported occurrence of each species in the GBIF database, climate data from the closet grid point was retrieved and added to the data set. If two reported occurrences were equally close to the same grid point, that point was only used once to avoid pseudo-replication. Once this data set was generated for each species, the 5^th, 50^th, and 95^th percentiles of each climate variable (see list of variables in supplement) were calculated and used to define the climate envelope of each species. Because we hypothesized that hydraulic architecture would relate to temperature and precipitation of the climate of origin, we focused our investigation on climate variables that would capture these abiotic stressors: MAP, MAT, temperature of the wettest and driest quarters, and MAP of the warmest quarter.

Statistical Analysis

Scaling relationships were evaluated using the ‘sma’ function in the ‘smatr’ package to account for variability in both x and y variables. In cases where we expected variation between the x and y variables to be explained by a power function, we log10-transformed both axes. This included relationships between leaf dimensions (width, length) and vessel number (Price et al. 2007. Gleason et al. 2018), and between leaf dimensions (i.e., pathlength ~ leaf length) and conduit diameter (Anfodillo et al. 2006). Axes were also log10-transformed in cases where we wanted to test an expected linear or proportional relationship between x and y variables, e.g., the relationship between conduit diameter and cell wall thickness at a given buckling pressure (Brodribb and Holbrook 2005), as well as conduit diameter (or conduit number) between different vein orders (Gleason et al. 2018). Variables were first tested for differences between genera by including ‘genus’ in a model to test for differences between the coefficients – slope (hereafter “scaling coefficient”) and intercept. If no differences were found, then a single scaling relationship was used and reported. If differences were found, then the genera with unique coefficients were removed and analyses were performed individually on each group.