Contrasts among cationic phytochemical landscapes in the southern United States

Abstract Understanding the phytochemical landscapes of essential and nonessential chemical elements to plants provides an opportunity to better link biogeochemical cycles to trophic ecology. We investigated the formation and regulation of the cationic phytochemical landscapes of four key elements for biota: Ca, Mg, K, and Na. We collected aboveground tissues of plants in Atriplex, Helianthus, and Opuntia and adjacent soils from 51, 131, and 83 sites, respectively, across the southern United States. We determined the spatial variability of these cations in plants and soils. Also, we quantified the homeostasis coefficient for each cation and genus combination, by using mixed‐effect models, with spatially correlated random effects. Additionally, using random forest models, we modeled the influence of bioclimatic, soil, and spatial variables on plant cationic concentrations. Sodium variability and spatial autocorrelation were considerably greater than for Ca, Mg, or K. Calcium, Mg, and K exhibited strongly homeostatic patterns, in striking contrast to non‐homeostatic Na. Even so, climatic and soil variables explained a large proportion of plants' cationic concentrations. Essential elements (Ca, Mg, and K) appeared to be homeostatically regulated, which contrasted sharply with Na, a nonessential element for most plants. In addition, we provide evidence for the No‐Escape‐from‐Sodium hypothesis in real‐world ecosystems, indicating that plant Na concentrations tend to increase as substrate Na levels increase.

As an often biologically critical element, Na is distinctive. It is generally considered nonessential for most plants, yet it is a key and essential nutrient for animals and decomposers (Clay et al., 2014;Kaspari, 2020;Kaspari, 2021;Kronzucker et al., 2013). The distribution of Na across terrestrial habitats is exceptionally heterogeneous.
Sodium is considered nonessential for the development of most plants (Grigore et al., 2012;Kronzucker et al., 2013). Notable exceptions in which Na benefits development or performance include most halophytes (Cheeseman, 2015;Flowers & Colmer, 2008;Kanai & Sakai, 2021) in certain environmental conditions, including specific ranges of Na concentration in the substrate (Santiago-Rosario et al., 2021). Certain C 4 (photosynthesis via C 4 carbon fixation or the Hatch-Slack pathway) and crassulacean acid metabolism (CAM) plants benefit -at specific substrate concentrations -from slight increases in substrate Na (Furumoto et al., 2011;Subbarao et al., 2003). Some C 4 plant species in the families Amaranthaceae, Asteraceae, Brassicaceae, Cyperaceae, Fabaceae, Poaceae, Portulacaceae, and Solanaceae, among others, found at relatively low concentrations of substrate Na, benefit from slight increases in Na by increasing biomass yield and reducing chlorosis (Brownell & Crossland, 1972;Johnston et al., 1988;Pessarakli & Marcum, 2000). For example, for species in the genus Flaveria (Asteraceae), Na is an essential nutrient as a transporter required for C 4 photosynthesis (Furumoto et al., 2011). Additionally, Na increased growth in the CAM species Bryophyllum delagoense (Crassulaceae) when substrate Na was increased to 0.1 meq/L NaCl as compared to individuals in basal culture solution (0.07 μeq/L NaCl), especially when grown under conditions of short-day length and high diurnal temperature variation (Brownell & Crossland, 1974). Therefore, at certain low concentrations of substrate Na, slight increases in Na appear to positively influence some C 4 and CAM plants' growth (Subbarao et al., 1999;Subbarao et al., 2003). However, it is important to note that whether Na′s effect on these species results from drought adaptations or metabolic micronutrient functions remains unresolved among plant physiologists (Brownell, 1968;Subbarao et al., 2003).
The No-Escape-from-Sodium hypothesis posits that plants' tissues broadly increase in Na concentration as the concentration of Na in the substrate or solution increases, irrespective of their growth responses, and there is empirical support for this pattern across selected plant taxa (Santiago-Rosario et al., 2021). However, our understanding of how plants respond to increasing substrate Na comes mostly from controlled laboratory and greenhouse experiments, which may or may not align with patterns in real-world ecosystems.
In the current study, we tested the No-Escape-from-Sodium hypothesis in the field across the southern continental United States. We also included three additional cations: calcium (Ca), magnesium (Mg), and potassium (K), because of the essential role they play in plant physiology and ecosystem processes and their ubiquitous distribution across the soilscape. We aimed to characterize the phytocationic landscapes for these four cations, by identifying the potential environmental drivers of plant cation concentrations and asking whether there is evidence for homeostatic regulation.
We collected Atriplex, Helianthus, and Opuntia aboveground photosynthetic tissues in 51, 131, and 83 sites, respectively. Collection sites across the southern U.S. ranged from Florida (~86° W) to California (~123° W). Sampling was completed during the summers of 2018, 2019, and 2020 ( Figure 1). We collected samples of aboveground tissues (leaves and stems), adjacent soil (top 10 cm), and one voucher specimen (for genus verification) from each site, along with GPS coordinates.
Plant and soil samples were oven-dried at 65°C for 7 days.
Aboveground tissues (i.e., leaves for Atriplex and Helianthus, and cladodes for Opuntia) were processed and ground into a fine, homogeneous powder for each site. Dried soil samples were passed through a 2 mm copper sieve to remove rocks and organic debris.
Concentrations of Ca, Mg, K, Na, and P on a dry mass basis for soils and plant tissues were determined at the Soil Testing and Plant Analysis Laboratory at Louisiana State University (http://www.lsuag center.com), using inductively coupled plasma with atomic emission spectrometry (ICP-AES) following standard protocols (Munns et al., 2010). Soil pH (1:1 water) was also measured. F I G U R E 1 Geographic locations and aboveground phytochemical landscapes of (a) Ca, (b) K, (c) Mg, and (d) Na (log 10 ppm) across the southern United States. Triangle, square, and circle shapes depict sites where Atriplex, Helianthus, and Opuntia were sampled, respectively. A color gradient demonstrates plant Ca, K, Mg, and Na concentrations, with darker shades indicating higher concentrations and lighter shades indicating lower concentrations.

| Abiotic conditions
Bioclimatic and elevation data were extracted for each sample site using the 'raster' package (Hijmans & van Etten, 2022) with a resolution of 4.6 km 2 in R (R Core Team, 2020). Climatic variables were mean annual temperature (MAT, °C), mean diurnal temperature range, temperature seasonality, annual precipitation (mm), precipitation in the wettest month (mm), precipitation in the driest month (mm), and precipitation seasonality (coefficient of variation).
We also measured the distance from each site to its nearest relevant coast (km), as a proxy for its proximity to its effective marine source of cations (as in Bravo & Harms, 2017). We used the Gulf of Mexico-Pacific Continental Divide, as wind movement and precipitation for both sides of the divide are associated more closely with their respective oceanic sources (Adams & Comrie, 1997). For each sample location, we expanded a circle using the Google Earth (http://www.google.com/earth) circumference tool until the edge of the circle first contacted the relevant coast. The radius of the circle was recorded as the effective distance to the nearest marine source for cations.

| Cation variability and spatial autocorrelation analysis
We performed a paired t-test to compare cation concentrations in aboveground plant tissues and adjacent soils and calculated the coefficient of variation to represent variation across space. To quantify whether inter-site proximity influenced cation concentration similarities for each element for aboveground plant tissues in each genus, as well as adjacent soils, we performed a Mantel test using the package 'ecodist' (Goslee & Urban, 2007). All cation concentrations were log 10 transformed and compared from site to site in a pairwise manner. All values were compiled into a matrix of differences (i.e., values closer to zero indicate more significant similarity in cation composition), and the absolute value was calculated to remove all negative values from the matrix. Additionally, we calculated the Haversine distance (km) among sites using the package 'geosphere' (Hijmans et al., 2019).
We performed Mantel tests to assess whether: (1) plants nearby shared similar cation concentrations, and (2) soils nearby shared similar cation concentrations. For a Mantel test, a significant result reveals that distances between two matrices are correlated (Rossi, 1996). Correlations can be positive or negative, representing how the variables are associated. Correlation in the 'ecodist' package is calculated using a Spearman approach, and all Mantel test calculations were performed using 9999 permutations. We also conducted a Mantel correlogram using the R function 'mgram' of the package 'ecodist' by dividing the data into 20 distance classes. For each of the three types of tests (1 and 2 above) a Mantel correlogram was performed for each cation and genus combination.

| Abiotic drivers of plant cation concentrations
To quantify the relative potential importance of abiotic conditions in determining plant cation concentrations, we fitted spatial regression models for each combination of plant genus and cation (12 models = 3 plant genera x 4 plant cations) with the Random Forest algorithm using the packages 'spatialRF' and 'ranger' (Benito, 2021;Wright & Ziegler, 2017). For each combination, we considered the following explanatory variables: MAT, temperature seasonality, mean annual precipitation, precipitation in the wettest months, precipitation in the driest months, precipitation seasonality, effective distance to coast, elevation, soil pH, and soil cation (i.e., the same as the plant cation). As multi-collinearity may affect the interpretation of variable importance of random forest models (Strobl et al., 2007), we estimated variance inflation (VIF) and correlation among abiotic variables and subsequently selected those whose VIF was lower than 4 and whose correlation with other abiotic variables was less than 0.7 (Dormann et al., 2013). We then tested the residuals of a non-spatial random forest model for spatial autocorrelation using multiscale Moran's I, and, if spatial autocorrelation was statistically significant, fitted spatial Random Forest models with Moran's eigenvector maps to build spatial predictors using the R function 'rf_spatial'. Because the default hyperparameters may not be adequate for each dataset, we used the R function 'rf_tuning' to select the optimal values for the number of regression trees in the forest, the number of variables to choose from on each tree split, and a minimum number of cases on a terminal node for each model. We repeated each model 50 times using the optimal hyperparameter values with the R function 'rf_repeat', because Random Forest is a stochastic algorithm whose variability may influence the interpretation of variable importance scores and response curves. We used median R-squared, calculated as the squared correlation between observed and predicted values, and normalized root mean square errors (NRMSE) to assess model fit across the 50 iterations of each model. We log 10 transformed plant and soil Na concentrations to normalize their distributions and only fitted a spatial random forest model for Opuntia and Na, since that was the only genus and cation combination for which it was required for our Random Forest analyses.

| Assessment of homeostasis
To assess the potential for plants to regulate cations in the field, we calculated the homeostasis coefficient Η (eta) for each cation and genus combination, as outlined in Sterner and Elser (2002). For this approach, we calculated a stoichiometric ratio based on phosphorus (P) concentration for plants and soils. Then the homeostasis coefficient (H) was calculated using the modified formula: where y is the plant stoichiometric cationic ratio, x is the cationic soil ratio, and c is a constant. By plotting the logarithms of the plant versus log y = log c + log x H soil stoichiometry, Sterner and Elser (2002) advise that slopes (1/H) between 1 and 0 indicate a continuum in homeostatic regulation with a slope value of 0 indicating strict homeostasis and a value of 1 indicating a lack of homeostasis between plant stoichiometry relative to soil stoichiometry (Meunier et al., 2014). However, as cations may be spatially autocorrelated, we accounted for any spatial autocorrelation using the package 'spaMM' (Rouseet & Ferdy, 2014). As a statistical approach, 'spaMM' fits mixed-effect models and permits the inclusion of spatial autocorrelation (i.e., Matern) as a random effect. The models generated in this study considered plant cation (denominated as X) stoichiometry (X plant :P plant ) and soil cation stoichiometry (X soil :P soil ) as fixed effects, and spatial autocorrelation was specified as a Matern random effect. For all models, we included spatial autocorrelation as a random effect, and fitted models using a nu of 0.5 to keep constancy across the cations and genera considered. All analyses were performed in R Studio (R Core Team, 2020).

| Phytochemical landscapes differ across cations
Aboveground

| Spatial autocorrelation is generally stronger for plant Na than the other cations across all genera
Aboveground plant Ca, Mg, and K exhibited generally low spatial autocorrelation across genera, especially focusing on positive spatial autocorrelation at distances <1000 km ( Figure 3). Atriplex had weak overall autocorrelation for plant Ca concentrations (r = 0.094, p = .043), and somewhat stronger autocorrelation for K concentrations (r = 0.154, p = .004; Table 2). Helianthus also had weak overall autocorrelation in K concentrations (r = 0.054, p = .050; Table 2).
In contrast, there was strong significant overall autocorrelation for plant Na concentrations across all three genera (p < .0001; Table 2).
On average, the correlograms for Na were decreasing, with positive significant spatial autocorrelation found for distances up to ~482 km for Atriplex, ~432 km for Helianthus, and ~ 162 km for Opuntia.

| Homeostasis coefficients vary across cations and taxa
The homeostasis coefficient for Na followed a different pattern than Ca, Mg, and K, with consistently low values for all genera.

| Soil cations and abiotic variables explain most of the variation in plant cation concentrations
Not all abiotic variables displayed the same influence on plant cation concentrations across the phytocationic landscapes ( Figure 5). All models for Atriplex, Helianthus, and Opuntia shared MAT and soil cation concentration as the highest contributing variables influencing aboveground plant cation concentrations, albeit at different levels of importance ( Figure 5, Table 4). Additionally, median R 2 values for all models ranged from 0.57-0.94 (see Table 4), thus representing a high level of explained variation in plant cation concentrations.
Mean annual temperature appeared to influence plant cation con-  (Figure 7). We do not describe the response curves for other variables because they were not shared among all genera due to multicollinearity (Table 4).

| DISCUSS ION
Plants across real-world ecosystems in the southern United States, on average, follow the No-Escape-from-Sodium hypothesis when exposed to variation in environmental Na across their ranges. This finding is congruent with patterns observed in laboratory stud-  (Table 3).

| Plant cation concentrations depend on the environment
Sodium is an unusual biotically important cation for plants as there is no apparent metabolic or structural function known for most plants (Benito et al., 2014;Kronzucker et al., 2013;Pardo & Quintero, 2002). for all genera, with Helianthus and Opuntia having higher variation in plant tissues than soil Na concentrations (Figure 3d). Variation in plant tissues of Atriplex was slightly lower than soil Na concentrations, which might reflect the halophytic nature of the family Amaranthaceae and the use of Na as a possible osmoticum in this genus (Glenn et al., 1994;White et al., 2017

| The phytochemical landscapes of Ca, Mg, and K appear to be at least partially controlled by homeostasis
Homeostasis plays a prominent role in some phytocationic landscapes, and responses differ substantially among genera (Wang et al., 2019). Calcium, Mg, and K were found at higher concentrations in plant tissues than in adjacent soils across all genera sampled.
Plants apparently regulate these cations because of their essential biochemical and physiological functions (see Table 5). The variation of these cations in plant tissues differs substantially from Na; variation in Ca, Mg, and K is extremely low (coefficient of variation approximately 5% in all three genera; Figures 1 and 2). Although Ca, Mg, and K displayed higher homeostatic levels than Na, the concentrations varied among genera (Gilroy et al., 1993;Leigh, 2001; Tang F I G U R E 5 Importance values of potential abiotic drivers of plant concentrations of (a) Ca (b) K (c) Mg and (d) Na for Atriplex, Helianthus, and Opuntia. Abiotic drivers are ranked by relative variable importance calculated for each random forest model; variable importance represents the increase in mean error across trees when a predictor is permuted. Black points and error bars are mean and 95% confidence values of relative importance values and were calculated across 50 iterations of the same random forest model. For visualization purposes, the relative importance scores of spatial predictors were excluded.  Sterner and Elser's (2002) definition), and Mg and K appeared to be kept at high levels of homeostasis in Atriplex (Table 3).

TA B L E 4 Model selection of environmental variable influences on plant cation concentrations obtained from the Random Forest analyses
For the genera mentioned above, soil Ca, Mg, and K concentrations do not substantially influence plant concentrations of the same ions other than being the sole or primary source of the element (Farago, 1994).
Homeostasis might explain why, on average, the importance of the climatic variables considered in this study differed considerably in explaining the variation in plant tissue Ca, Mg, and K across the geographic range sampled (Figure 6, Tables 3 and 4). The role that Ca, Mg, and K play in plants varies greatly, from a fundamental structural component to metabolism and enzymatic reactions (Table 5).
Because of anthropogenic global warming and increases in atmospheric CO 2 concentrations, plant quality and stoichiometric mechanisms might be highly impacted, especially in highly regulated elements such as Ca and Mg. Evidence suggests warming conditions and increasing CO 2 concentrations will affect the environmental availability of some essential elements and their regulation by plants (i.e., N and P) unevenly across their ranges (Dijkstra et al., 2012;Gu et al., 2017). Moreover, the nutrition dilution hypothesis posits that increases in atmospheric CO 2 , water availability, and temperature promote increased carbohydrate production in primary producers resulting in increases in plant biomass accumulation with low foliar nutrient quality, which in turn promotes a decline in herbivore abundance (Welti et al., 2020). Whether these stoichiometric patterns hold across other micronutrients and nonessential elements across plant taxa and different habitats remains unquantified.
Among plants' essential elements, such as Ca, Mg, and K, host-specific herbivores should encounter a relatively consistent F I G U R E 6 Responses of plant concentrations of (a) Ca, (b) Mg, (c) K and (d) Na for Atriplex, Helianthus, and Opuntia to mean annual temperature (MAT). Each line represents a response curve estimated by a random forest model, which was fitted for each combination of plant genus and cation and repeated 50 times. Plant Na is on a log 10 scale.
concentration of these elements across their ranges, yet no such patterns were observed for Na. Na is an essential element for  (Boggs & Dau, 2004;Bravo et al., 2010;Burger & Gochfeld, 2003;Clay et al., 2017;Holdø et al., 2002), whereas high levels of plant Na concentrations have generated unique herbivore adaptations to prevent salt-induced stress encountered in some halophytic plant taxa that accumulate Na in plant tissues to evade herbivory (Kenagy, 1973;Renault et al., 2016). Yet, the mechanisms by which Na variation influences animal behavior across natural settings remain incompletely studied, especially when considering herbivorous species with large ranges.

| Plants share similar Na concentrations the closer they are to each other
Sodium's high variability differs geographically, especially across soil Na gradients and proximity to coastlines with persistent salt deposits from marine aerosols (Borer et al., 2019;Bravo & Harms, 2017;Doughty et al., 2016). Not surprisingly, plant Na concentration exhibited strong spatial autocorrelation across all genera sampled, emphasizing the weak homeostatic regulation plants have for this cation along with the environmental influence on plant Na acquisition ( Figure 3, Table 2). Individuals closer to each other share similar levels of tissue Na across a heterogeneous landscape. Moreover, Ca, Mg, and K, patterns of spatial autocorrelation were complex, albeit mostly weak across genera (Table 2). Potassium showed modest spatial autocorrelation only in Atriplex (Table 2). A similar pattern was observed across several plant families sampled geographically broadly in China, where K and Na showed strong spatial F I G U R E 7 Responses of plant concentrations of (a) Ca, (b) Mg, (c) K and (d) Na for Atriplex, Helianthus, and Opuntia to soil cation concentrations (ppm). Each line represents a response curve estimated by a random forest model, which was fitted for each combination of plant genus and cation and repeated 50 times. Plant and soil Na are on a log 10 scale. autocorrelation across leaf tissues, but Ca did not, thus suggesting that these general patterns might be shared globally across plant taxa (Zhang et al., 2012).

| CON CLUS ION
Our study illustrates the utility of focusing attention on the formation and maintenance of the phytochemical landscapes of essential (i.e., Ca, Mg, and K) and generally nonessential elements to plants, such as Na.
The No-Escape-from-Sodium hypothesis was supported across fieldcollected plants, which suggests that plant tissue concentrations tend to reflect Na in the substrate, similar to patterns observed in controlled settings (Santiago-Rosario et al., 2021). The differences in cation variation and spatial autocorrelation observed in this study appear to be linked to homeostatic regulation, or lack thereof, depending on elemental essentiality to plants. Thus, identifying general phytochemical patterns of essential and nonessential elements to plants across the landscape represents a key step toward better understanding biogeochemical cycles and their effects on trophic-level interactions and ecosystem dynamics (Hunter, 2016;Sterner & Elser, 2002). Expanding this type of research to other essential and nonessential elements, other taxa, and additional geographic locations would broaden our understanding of the evolutionary and biogeographic processes that give rise to phytocationic landscapes.

ACK N OWLED G M ENTS
Texas Ecological Laboratory and private Texas landowners supported and funded this work and graciously gave access to collect on their properties during the summer of 2020. We thank Dr. Katherine Hovanes, Colin Morrison, Diego Paredes-Burneo, and Juan A.
Santiago Rosario for their help and support during field collections.

This study was supported by the Texas Ecological Laboratory
Program.

CO N FLI C T S O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in Dryad at https://doi.org/10.5061/dryad.2bvq8 3bs2.