Introduction
The seed coat is a physical barrier that protects the embryo by deterring pathogens (Mohamedyasseen et al. Reference Mohamedyasseen, Barringer, Splittstoesser and Costanza1994) and limiting expulsion of chemical cues used by adversarial organisms (Paulsen et al. Reference Paulsen, Colville, Kranner, Daws, Hogstedt, Vandvik and Thompson2013). The structural integrity of the seed coat influences weed seed longevity in soil (Davis et al. Reference Davis, Schutte, Iannuzzi and Renner2008; Gardarin et al. Reference Gardarin, Durr, Mannino, Busset and Colbach2010) and is a target for management interventions that utilize biological control (Muller-Stover et al. Reference Muller-Stover, Nybroe, Baraibar, Loddo, Eizenberg, French, Sonderskov, Neve, Peltzer, Maczey and Christensen2016) or mechanical devices (Walsh et al. Reference Walsh, Broster, Schwartz-Lazaro, Norsworthy, Davis, Tidemann, Beckie, Lyon, Soni, Neve and Bagavathiannan2018b) to lessen the amount of weed seeds in soil. To facilitate understanding of conditions and tactics that fracture weed seed coats, in this study we introduce a colorimetric assay for quickly (<32 h) distinguishing damaged from nondamaged seeds of different weed species.
Rapid tests for seed quality include assessments of seed exudates (Elias et al. Reference Elias, Copeland, McDonald and Baalbaki2012). Exudates are indicative of seed quality because damaged seeds incur internal injury during imbibition and expel large amounts of material into the imbibition media (Pandey Reference Pandey, Linskens and Jackson1992). Exudates from damaged seeds include hydrogen ions (H+) and compounds that generate H+ in solution. Therefore, low pH in seed exudate is a sign of a seed that might be nonviable or damaged (Elias et al. Reference Elias, Copeland, McDonald and Baalbaki2012; Verma et al. Reference Verma, Verma and Tomer2003).
Resazurin is a water-soluble dye that changes colors and visible light absorbance properties as the pH of the solution decreases below 6.8. These pH-induced color transitions are caused by the reduction of resazurin to resorufin, which changes the solution from blue to pink and its light absorbance maximum from λ = 600 nm to λ = 570 nm (Guerin et al. Reference Guerin, Mondido, McClenn and Peasley2001; Twigg Reference Twigg1945). As pH further declines, resorufin is reduced to dihydroresorufin, which causes a change from pink to colorless (Guerin et al. Reference Guerin, Mondido, McClenn and Peasley2001; Twigg Reference Twigg1945). In addition to being a colorimetric indicator for pH, resazurin is also a marker for cellular metabolic activity because living cells in culture absorb resazurin, then reduce resazurin to resorufin and excrete resorufin into the extracellular medium (O’Brien et al. Reference O’Brien, Wilson, Orton and Pognan2000). As a metabolic indicator, resazurin was used to identify individual seeds of radish (Raphanus sativus L.) that were subjected to artificial ageing (Min and Kang Reference Min and Kang2011). Sensitivity to chemical changes caused by individual seeds (e.g., Min and Kang Reference Min and Kang2011), combined with resazurin reactions to decreasing pH (Twigg Reference Twigg1945) and expectations for effluxes of H+ from mechanically damaged seeds (Elias et al. Reference Elias, Copeland, McDonald and Baalbaki2012) suggest that resazurin can be used to identify individual weed seeds with physical injury to the seed coat and cell membranes. However, resazurin responses to seeds that differ in structural integrity have not been determined. The aim of this study was to determine whether mechanically damaged weed seeds can be distinguished from nondamaged weed seeds by using a resazurin solution to characterize individual seed steepates.
Materials and Methods
Plant Materials
Study species included barnyardgrass, curly dock, junglerice, kochia, oakleaf datura, Palmer amaranth, spurred anoda, stinkgrass, tall morningglory, and yellow foxtail. This set of species comprised grass weeds (barnyardgrass, junglerice, stinkgrass, yellow foxtail), broadleaf weeds with water-permeable seed coats (curly dock, kochia, Palmer amaranth), and broadleaf weeds with water-impermeable seed coats (oakleaf datura, spurred anoda, tall morningglory; Baskin and Baskin Reference Baskin and Baskin2014). Furthermore, the set of study species included weeds that differed in seed size. Mean masses of 100-seed samples (N = 3) were as follows: tall morningglory, 2.68 g; oakleaf datura, 1.07 g; spurred anoda, 0.91 g; curly dock, 0.26 g; yellow foxtail, 0.20 g; barnyardgrass, 0.18 g; junglerice, 0.13 g; kochia, 0.08 g; Palmer amaranth, 0.04 g; stinkgrass, 0.03 g.
For each study species, natural dispersal units (herein “seeds”) were collected from populations occurring in agricultural fields at the New Mexico State University, Leyendecker Plant Science Research Center (32.19°N, 106.74°W). The year in which seeds were collected differed among species, with oakleaf datura, tall morningglory, spurred anoda, Palmer amaranth, barnyardgrass, and stinkgrass collected in 2017; yellow foxtail and junglerice collected in 2015; kochia collected in 2012; and curly dock collected in 2009. Seeds were harvested from plants by hand, air-dried at 25 C for 14 to 20 d, cleaned using an air-column separator, and stored at 4 C in airtight containers. For broadleaf weeds including curly dock, oakleaf datura, Palmer amaranth, spurred anoda, and tall morningglory, seeds did not contain fruit tissues or accessory structures. For kochia, seeds were single-seeded fruits with thin, unattached pericarps (utricles). For grass weeds including barnyardgrass, junglerice, and yellow foxtail, seeds were caryopsis with lemma, palea, and glumes. For stinkgrass, seeds were caryopsis with lemma. Just prior to use in the study, seeds were assayed for viability using a 0.6% aqueous solution of 2,3,5-triphenyl-tetrazolium chloride (Peters Reference Peters2000). For each species, initial seed viability was determined to be high (>90%).
Seed Testing
Seeds were subjected to four mechanical damage treatments: 1) abrasion, 2) piercing, 3) slicing, and 4) grinding. Nondamaged seeds served as controls. An experimental run was 48 seeds for each combination of species and seed damage type, which resulted in a total of 2,400 seeds run−1. The study included four experimental runs, with successive runs offset by 2 wk. The combined mechanical damage treatments represented a range of damage severities (Figure 1). Abrasion and piercing subjected seeds to small amounts of mechanical damage, whereas slicing and grinding severely damaged seeds. Damage treatments in this study produced seed injuries consistent with those obtained with impact mills that pulverize or abrade weed seeds collected with chaff during mechanical harvest of grain crops (Berry et al. Reference Berry, Fielke and Saunders2014; Walsh et al. Reference Walsh, Broster and Powles2018a).
Figure 1. Seed damage treatments exemplified with spurred anoda (A through E) and yellow foxtail (F through J). Seed damage treatments were nondamaged (A, F), abraded (B, G), pierced (C, H), ground (D, I), and sliced (E, J).
Seeds were abraded with a custom-made, hand-operated scarifier featuring a stationary platform (27.9 cm × 25.4 cm) and detached block (17.8 cm × 12.7 cm × 2.5 cm) that were each coated with medium-grit sandpaper (Coated Abrasive Manufacturers Institute grit designation 60). Batches of seeds were placed on the scarifier platform and rubbed with firm pressure using the detached block. Preliminary observations indicated that if rub duration and seed batch size were consistent across species, the scarifier excessively pulverized seeds of small-seeded species while not abrading seeds of larger-seeded species. Accordingly, numbers of seeds per batch and rub durations were unique to species. For tall morningglory, oakleaf datura, and spurred anoda, scarifier operation parameters were 50 seeds batch−1, 15 s batch−1. For kochia, yellow foxtail, barnyardgrass, and junglerice, scarifier operation parameters were 100 seeds batch−1, 10 s batch−1. For curly dock, Palmer amaranth, and stinkgrass, scarifier operation parameters were 150 seeds batch−1, 20 s batch−1. Seed piercing was performed under a microscope using a stainless steel dissecting needle. Due to their minute size, stinkgrass seeds could not be pierced. Palmer amaranth seeds could not be pierced because Palmer amaranth seed coats shattered during piercing. Seeds were sliced using a single-edged razor blade under a dissecting microscope attached with a dual goose-neck fiber-optic illuminator. Seed grinding procedures differed among species. These species-specific procedures enabled similar degrees of damage across species that varied in seed size. For tall morningglory, oakleaf datura, and spurred anoda, seeds were individually ground using a metallic mortar and pestle, with each seed struck once with the pestle. For kochia, yellow foxtail, barnyardgrass, junglerice, curly dock, Palmer amaranth, and stinkgrass, seeds were individually ground inside the wells of 96-well plates using a blunt metallic rod.
Mechanically damaged and nondamaged seeds were individually steeped in deionized water using 96-well plates (1 seed well−1). Volumes of water added to wells were minimal amounts of water needed to create a recoverable solution following immersion of the seed. For tall morningglory and oakleaf datura, 250 µL of water was added to individual wells. For spurred anoda, 200 µL of water was added to individual wells. For curly dock, kochia, yellow foxtail, and junglerice, 100 µL of water was added to individual wells. For Palmer amaranth, and stinkgrass, 75 µL of water was added to individual wells. Well-plates with seeds and water were covered with aluminum foil and transferred to an incubator set to 30 C. Plates were incubated for 12 h. Preliminary observations indicated that this incubation period was optimum because it both maximized exudate concentrations in steepates and prevented microbial contamination that would otherwise interfere with resazurin reactions. Following incubation for 12 h, well-plates were centrifuged at 4,000 g for 10 min, and then 50 µL of steepate from each well was transferred to the respective well of a new well-plate for the colorimetric assay.
Colorimetric assays were performed using a 0.1% aqueous solution of resazurin (7-hydroxy-3H-phenoxazin-3-one-10-oxide) sodium salt. Two microliters of this resazurin solution was added to each 50-µL steepate sample in the 96-well plates. Steepate-resazurin solutions were mixed with pipette tips and immediately assessed for light absorbance at λ = 600 nm (herein, light absorbance at λ = 600 nm is abbreviated “A600”) using a microplate spectrophotometer (SpectraMax 190 Microplate Reader, Molecular Devices, LLC, 3860 N First Street, San Jose, CA 95134). Well-plates were then covered with aluminum foil and incubated at 30 C. Color changes over time in the steepate-resazurin solutions were determined by measuring A600 at 2 h, 5 h, 20 h, and 40 h after the initial mixing (0 h) of steepate and resazurin solutions. Hereafter, measurement intervals for steepate-resazurin solutions are referred to as “reaction durations.”
Data Analysis
All statistical analyses were performed using the open source statistical software program R (v.3.6.1, The R Foundation for Statistical Computing, http://www.r-project.org). Levene’s tests for homogeneity of variances performed with the R library car indicated equal variances among runs. Accordingly, data from runs were combined.
Data were analyzed with a sequence of three steps. The first step determined species-specific reaction durations that first produced differences in A600 between nondamaged seeds and all damaged seeds. The second data analysis step developed classification trees for distinguishing nondamaged and damaged seeds based on A600 values. The third data analysis step evaluated classification trees by inspecting accuracies and inaccuracies of model predictions. We wanted to know whether the performance of the resazurin assay was contingent on seed damage type and species. Accordingly, we developed classification trees for each combination of species, damage type, and experimental run; and we statistically analyzed a measure of model accuracy to identify possible species and damage type effects on classification tree performance.
To determine minimum resazurin reaction durations needed for differences in A600 between nondamaged and damaged seeds of the same species, we performed ANOVAs and post hoc Tukey honestly significant difference (HSD) tests to identify differences in A600 among seed damage types. For these analyses, steepates derived from individual seeds were considered subsamples. Accordingly, data for ANOVAs and post hoc Tukey HSD tests were averages across the individuals within an experimental run, species, damage type, and reaction duration.
Classification trees for the binary response variable “nondamaged” or “damaged” were developed for each combination of species, damage type, and experimental run using the R library rpart (Crawley Reference Crawley2013). Predictor variables in these classification trees were A600 values from individual seeds for the reaction duration when nondamaged and damaged seeds first differed with respect to A600 (aforementioned minimum resazurin reaction durations determined with ANOVAs and post hoc Tukey HSD tests). Classification tree size was constrained to one node. Therefore, classification trees were dichotomous decision trees featuring a single “if-then” condition in A600 that predicted whether a seed was damaged or nondamaged.
We assessed classification trees by determining prediction accuracy rates, which were the percentages of all seeds correctly predicted with the decision rules. Prediction accuracy rates were assessed for interactions between species and damage type by comparing two linear mixed effects (LME) models. One LME model featured fixed effects for species, damage type, and the interaction between species and damage type, and one LME model had fixed effects for species and damage type only. For each LME, the random effect was “experimental run” and maximum likelihood was used for fitting. The R library lme4 was used for model fitting. To meet assumptions of parametric analysis, prediction accuracy rates were subjected to the arcsine square root transformation prior to the analysis. Once developed, the two LME models were compared with the likelihood ratio test using the anova function in R (Pinheiro and Bates Reference Pinheiro and Bates2000).
To increase understanding of general causes for incorrect predictions, we calculated two metrics for prediction inaccuracy: 1) false negative rate, which was the percentage of seeds predicted to be nondamaged when actually they were damaged; and 2) false positive rate, which was the percentage of seeds predicted to be damaged when actually they were nondamaged. These metrics for inaccuracy were calculated for each classification tree. A two-tailed, paired sample Wilcoxon rank test was used to compare false negative rates against false positive rates.
Results and Discussion
A600 values for nondamaged seeds were greater than A600 values for damaged seeds for all broadleaf weeds except kochia (Table 1). For broadleaf weeds including tall morningglory and curly dock, differences in A600 between nondamaged and damaged seeds were apparent after 2 h of the resazurin reaction. Following 5 h of the resazurin reaction, A600 values differed between nondamaged and damaged seeds for spurred anoda. Twenty hours of resazurin reaction was required for differences in A600 between nondamaged and damaged seeds of Palmer amaranth and oakleaf datura.
Table 1. Absorbance at 600 nm (A600) for resazurin reactions that utilized steepates from seeds subjected to mechanical damage treatments.a
a Data are means of four replications, with each replication comprised of 48 individual seeds. Within a row (combination of species and reaction duration), means with the same letter are not significantly different according to Tukey’s honestly significant difference test (α = 0.05).
b Abrading with sandpaper.
c Grinding with mortar and pestle or blunt metallic rod.
d Piercing with a sharp needle.
e Slicing with a scalpel.
f The pierced treatment was not executed on Palmer amaranth and stinkgrass seeds.
For grass weeds, the resazurin reaction revealed differences between nondamaged seeds and only certain types of damaged seeds. For ground, pierced, and sliced seeds of barnyardgrass and yellow foxtail, A600 values were less than those for nondamaged seeds; however, A600 values for abraded seeds were similar to those for nondamaged seeds. For stinkgrass, A600 values were similar between nondamaged and abraded seeds, but they differed between nondamaged and ground seeds, and between nondamaged and sliced seeds. Nondamaged seeds of junglerice produced A600 values that differed from those of both ground and sliced seeds, but A600 values for nondamaged junglerice seeds were similar to the A600 values for abraded and pierced seeds.
Significant differences in A600 between nondamaged and damaged seeds generally coincided with variability in resazurin-steepate solution color (Supplemental Figure S1). Nondamaged seeds tended to produce resazurin-steepate solutions that remained blue for up to 20 h, whereas resazurin-steepate solutions for damaged seeds changed from blue to pink. Such variation in resazurin-steepate solution color, combined with differences in A600, indicated that damaged seeds induced the reduction of resazurin to resorufin within 0 to 20 h of the resazurin reaction; but within this timeframe, relatively few nondamaged seeds altered steepates in ways that were detectable with the resazurin assay. For some combinations of species and damage type, resazurin-steepate solutions changed from blue to faint orange (e.g., ground seeds of tall morningglory) or from blue to pink to colorless (e.g., ground seeds of oakleaf datura). These orange and clear solutions indicated that resazurin was reduced to resorufin, which was subsequently reduced to dihydroresorufin (Guerin et al. Reference Guerin, Mondido, McClenn and Peasley2001; Min and Kang Reference Min and Kang2011). Dihydroresorufin can be reoxidized to resorufin by atmospheric oxygen, but atmospheric oxygen cannot reoxidize resorufin to resazurin (Guerin et al. Reference Guerin, Mondido, McClenn and Peasley2001; Twigg Reference Twigg1945). Accordingly, colorless steepate solutions reversibly changed to pink, but pink steepate solutions did not change back to blue. We occasionally observed colorless steepate solutions for nondamaged seeds of kochia and oakleaf datura. Although these nondamaged seeds of kochia and oakleaf datura were not intentionally damaged, they exuded materials that caused chemical reductions in the resazurin-steepate solutions.
Prediction accuracy rates for classification trees were influenced by an interaction between species and damage type (χ225 = 38.6, P = 0.04). For 22 of the 38 combinations of species and damage type, classification trees accurately predicted binary damage states (nondamaged or damaged) for at least 90% of seeds (Table 2). Classification tree predictions were at least 95% accurate for 11 of the 38 combinations of species and damage type. Inaccurate predictions were, in general, more associated with nondamaged seeds that were incorrectly classified as damaged rather than damaged seeds that were incorrectly classified as nondamaged. This is because classification trees generally produced sets of predictions with higher false positive rates (median = 8.5%) than false negative rates (median = 6.4%; paired sample Wilcoxon rank test, N = 152, Z = −2.71, P < 0.01).
Table 2. Overall accuracies for classification trees for distinguishing nondamaged and damaged seeds based on absorbance at 600 nm (A600) of individual seed steepates after mixing with resazurin solution.a
a Classification trees were developed separately for combinations of species and damage type (i.e., abraded, ground, pierced, sliced). For all species except kochia, predictor variables in classification trees were minimum durations for detecting differences between nondamaged and damaged seeds (Table 1). Data are means with standard errors in parentheses, N = 4.
b Percentage of seeds correctly predicted with decision rule in classification tree.
c Predictor variables in classification trees were A600 values at 20 h after the initial mixing of steepate and resazurin solutions.
d The pierced treatment was not executed on Palmer amaranth and stinkgrass seeds.
Several previous studies developed techniques for distinguishing weed seeds based on morphology and color (e.g., Arefi et al. Reference Arefi, Motlagh and Khoshroo2011; Granitto et al. Reference Granitto, Navone, Verdes and Ceccatto2002, Reference Granitto, Verdes and Ceccatto2005; Petersen and Krutz Reference Petersen and Krutz1992); however, to our knowledge, only two previous studies created procedures for sorting intact versus nonintact weed seeds: 1) Matzrafi et al. (Reference Matzrafi, Herrmann, Nansen, Kliper, Zait, Ignat, Siso, Rubin, Karnieli and Eizenberg2017), who used hyperspectral imaging to distinguish germinated from nongerminated seeds of Palmer amaranth; and 2) Schutte et al. (Reference Schutte, Haramoto and Davis2010), who used steepate conductivity to separate intact seeds from seeds with fissured coats of velvetleaf (Abutilon theophrasti Medik.) and ivyleaf morningglory (Ipomoea hederaceae Jacq.). For specific combinations of species and damage type, our resazurin assay was able to correctly classify more than 90% of seeds. These high prediction accuracy rates were equivalent to or exceeded prediction accuracy rates in previously developed techniques for sorting intact and nonintact weed seeds (Matzrafi et al. Reference Matzrafi, Herrmann, Nansen, Kliper, Zait, Ignat, Siso, Rubin, Karnieli and Eizenberg2017; Schutte et al. Reference Schutte, Haramoto and Davis2010).
The resazurin assay featured relatively low prediction accuracy rates for pierced and abraded seeds, especially abraded seeds of grass weeds. Prediction accuracy rates were high for ground seeds of specific broadleaf species and yellow foxtail, pierced seeds of curly dock, and sliced seeds of specific broadleaf and grass species. Species and damage types for which the resazurin assay performed well suggest that the assay predictions were improved by uninhibited flow of exudate from damaged seeds. Most notably, relatively large amounts of exudate might have been expelled from ground and sliced seeds because grinding and slicing subjected seeds to higher degrees of mechanical damage than abrading and piercing. Also, broadleaf weed seeds might have enriched steepates with exudate because seeds of broadleaf weeds featured fewer embryo-covering structures than seeds of grass weeds. The resulting higher concentrations of exudates enhanced reduction of resazurin to resorufin, thereby causing sharp declines in A600 for steepates of damaged seeds. To improve performance of the resazurin assay for pierced and abraded seeds, including abraded seeds of grass weeds, subsequent research should consider approaches for enhancing exudate flow from damaged seeds and promoting resazurin solution sensitivity to H+ exuded from damaged seeds. Such procedures may include gentle agitation to seed steepates during their formation and increased concentration of resazurin in resazurin-steepate solutions. However, increased resazurin concentrations may conflict with the aim for rapidity because our preliminary experiments indicated that resazurin concentrations greater than that used in this study (0.1% resazurin) prolonged the resazurin reaction in seed steepates.
In this study, kochia was the only species for which resazurin could not distinguish nondamaged seeds from any type of damaged seed. Previous studies determined that kochia embryo-covering structures were among the thinnest (Davis et al. Reference Davis, Schutte, Iannuzzi and Renner2008) and physically weakest (Davis et al. Reference Davis, Fu, Schutte, Berhow and Dalling2016) for collections of annual weed species that are common in North America. Additional previous studies determined that kochia seeds were not highly dormant (Friesen et al. Reference Friesen, Beckie, Warwick and Van Acker2009) and completed germination in less than 24 h when wetted under temperatures between 20 C and 35 C (Al-Ahmadi and Kafi Reference Al-Ahmadi and Kafi2007; Orlovsky et al. Reference Orlovsky, Japakova, Shulgina and Volis2011)—temperatures that are comparable to the ambient temperatures under which this study occurred. Rapid germination or thin, weak seed coats might have prevented the resazurin assay from distinguishing damaged from nondamaged seeds of kochia because such conditions would have promoted the release of exudates from nondamaged seeds. The possible mechanisms underlying the poor performance of the resazurin assay for kochia suggest that kochia is inherently ill-suited for the resazurin assay for identifying mechanically damaged weed seeds. Procedural modifications, such as those previously proposed for pierced and abraded seeds, are not likely to improve the performance of the resazurin assay for kochia.
Seeds subjected to mechanical damage are less likely to persist in soil compared with nondamaged seeds (Davis et al. Reference Davis, Schutte, Iannuzzi and Renner2008; Shergill et al. Reference Shergill, Kreshnik, Davis and Mirsky2020). Accordingly, the resazurin assay presented in this study might be used to quickly assess the efficacies of management interventions that target weed seeds during crop harvest (i.e., “harvest weed seed control” [HWSC]). HWSC technologies including the Harrington Seed Destructor and the integrated Harrington Seed Destructor are mills that pulverize or abrade weed seeds collected during mechanical harvest of grain crops (Walsh et al. Reference Walsh, Broster, Schwartz-Lazaro, Norsworthy, Davis, Tidemann, Beckie, Lyon, Soni, Neve and Bagavathiannan2018b). Current techniques for evaluating Harrington Seed Destructors include germination or seedling emergence assays that last 14 to 28 d (Berry et al. Reference Berry, Fielke and Saunders2014; Tidemann et al. Reference Tidemann, Hall, Harker and Beckie2017). Although germination and seedling emergence assays are effective for assessing Harrington Seed Destructor efficacy, they do not facilitate operational modifications that might be necessary when Harrington Seed Destructors are applied to new cropping systems and novel weed flora. We believe that tests for seed damage that are more rapid than germination and emergence assays can promote both innovation in HWSC technologies and expansion of Harrington Seed Destructor concepts to new crops. However, before the resazurin assay is used in HWSC assessment, additional research must confirm that seeds producing positive assay responses also fail to generate seedlings. This is because the resazurin assay detects damage to seed coats and does not assess capacities for germination and emergence.
In this investigation we introduced a colorimetric assay that utilizes common laboratory tools and basic principles of seed quality testing to identify mechanically damaged weed seeds in less than 32 h. Our resazurin assay is currently able to distinguish nondamaged seeds from seeds subjected to moderate-to-severe damage for specific broadleaf and grass weeds. We believe that the recommendations presented in this study will guide improvements that enable the resazurin assay to detect small amounts of damage on surfaces of seeds for many different weed species. However, we caution against using the resazurin assay for species featuring seeds with very thin seed coats or seeds with capacities for rapid germination. Such conditions are not amenable to the resazurin assay for distinguishing mechanically damaged weed seeds. Rapid assessments of seed intactness can accelerate the development of tactics for reducing the number of weed seeds in soil. Thus, we advocate further development and use of resazurin assays by laboratories studying weed seedbank depletion though HWSC.
Supplementary Material
To view supplementary material for this article, please visit https://doi.org/10.1017/wet.2019.125
Acknowledgements
This project was funded by USDA–National Institute of Food and Agriculture Hatch Project NM-Schutte-12H. Salaries and research support were provided by state and federal funds appropriated to the New Mexico Agricultural Experiment Station. No conflicts of interest have been declared. We gratefully acknowledge the assistance of Ms. Helen Vessels for photographic support, and Mr. Ed Morris for help with seed collection.
