Phenotypic plasticity in diaspore production of a amphi-basicarpic cold desert annual that produces polymorphic diaspores

Phenotypic plasticity has been studied in diaspore-dimorphic species, but no such study has been done on a diaspore-polymorphic species. Our aim was to determine the effects of abiotic and biotic factors on phenotypic plasticity of the diaspore-polymorphic cold desert annual Ceratocarpus arenarius. Plants produced from dispersal units near the soil surface (a, basicarps) and at the middle (c) and upper (f) parts of the plant canopy were subjected to different levels of soil moisture, nutrient supply and intramorph and intermorph densities. Different levels of these biotic and abiotic factors resulted in significant variation in total plant mass, diaspore mass, mass allocation to stem and reproductive organs and total number and proportion of morphs a, c and f on an individual. The effect of stress on number and mass of a dispersal unit morph varied by treatment, with dispersal unit f having the highest CV and dispersal unit a the lowest. The success of this diaspore polymorphic species in its rainfall-unpredictable environment likely is enhanced by plasticity in production of the different types of diaspores.


Results
Dry mass accumulation and allocation. Total plant dry mass and mass of reproductive organs (mass of all dispersal units) were significantly affected by different levels of all treatments in all harvested plants, except D 1 ( Fig. 2A) for plants derived from dispersal unit morph c, D 2(a:c) and D 2(a:f) (Fig. 2B) for plants from all dispersal unit morphs and D 3 (Fig. 2C) for plants derived from dispersal unit morph c (Fig. 3). Mass allocation to vegetative and reproductive organs was significantly affected by all treatments (Fig. 4). Allocation to reproductive organs was significantly higher in high than in low water treatment for all harvested plants derived from the three dispersal unit morphs (Fig. 4A). In the nutrient supply treatment, however, the highest allocation to reproductive organs occurred in different nutrient levels for plants derived from the three dispersal unit morphs (i.e. high, moderate and low nutrient supply in plants derived from dispersal unit morphs a, c and f, respectively) (Fig. 4B). In response to different levels of D 1 (Fig. 4C) and D 3 (Fig. 4E) treatments, allocation to reproductive organs was significantly higher in high than in low density stress for plants derived from the dispersal unit morph a but nonsignificantly higher in low than in high density stress for plants derived from the dispersal unit morph f. Moreover, allocation to reproductive organs in plants derived from dispersal unit morph c was similar www.nature.com/scientificreports/ to that of plants derived from dispersal unit morphs a and f in D 1 and D 3 , respectively. In response to different combinations and levels of the D 2 treatment, only [D 2(c:f) ] differed significantly in allocation to reproductive organs among the different density levels, which were higher in low than in high density stress for the plants derived from the dispersal unit morphs c and f (Fig. 4D).
Dispersal unit and fruit production. With increase in level of stress in which plants derived from the three dispersal unit morphs were grown, total number and total mass of the three dispersal unit morphs and number ( S4) for dispersal unit morphs a, c and f with a fruit were almost the same as that of each dispersal unit morph in response to different levels of all treatments.
In the water stress, nutrient availability, D 1 and D 3 treatments, two-way ANCOVAs showed that proportion of dispersal unit morphs c and f was significantly affected by water stress and nutrient availability, and that proportion of dispersal unit morph a was significantly affected by nutrient availability and D 1 (Table 1). Moreover, there were significant differences in proportion of dispersal unit morphs c and f among plants from the three dispersal unit morphs only in the D 1 treatment (Table 1). In the D 2 treatment, three-way ANCOVAs showed that only proportion of dispersal unit morphs c and f was significantly affected by combination of any two dispersal unit morphs (Table 2). Correlative analyses also showed that the relationship between total plant mass and proportion of dispersal unit morphs a and c was negatively correlated in most treatments (Supplementary  Tables S1-S5). However, the relationship between total plant mass and proportion of dispersal unit morph f was positive in most treatments (Supplementary Tables S1-S5).

Discussion
The most important findings of our study were that most levels of abiotic (soil water content and nutrient supply) and biotic (density) treatments caused a decrease in number and mass of the uppermost dispersal unit (morph f) of C. arenarius but an increase or no change in number and mass of the lowermost dispersal unit (morph a). Moreover, a striking result of the present study is that trends for number and mass of dispersal unit morph c were the same as those of dispersal unit morphs a or f, depending on level of treatment. Further, in most stress conditions there was a downward shift in proportion of mass and number for dispersal unit morph f, while proportion of number and mass for dispersal unit morph a increased. Thus, our hypothesis that increasing stress causes a decrease in production of dispersal units with high dispersal and low dormancy in favor of dispersal units with high dormancy and no dispersal is supported.
Both water and density treatments caused a significant shift in proportion of dry mass allocated to vegetative and reproductive components in the diaspore-dimorphic species Diptychocarpus strictus 18 and Lappula duplicicarpa 19 . In particular, the proportion of dry mass allocated to reproductive organs increased in stressed plants, while the proportion of dry mass allocated to vegetative organs in these two species and also in C. arenarius correspondingly decreased. In C. arenarius, however, variation in reproductive mass (CV 89.22%, 92.45% and www.nature.com/scientificreports/ 86.45% in plants derived from dispersal unit morphs a, c and f, respectively) was higher than that of total plant mass (CV 87.78%, 82.35% and 85.60%, respectively). The conditions to which the parent plant is exposed influence the number 16,23 , mass 15,36 and size 23,37 of diaspores produced. The proportion of diaspore morphs produced by heteromorphic plants [i.e. (HDi-LDo)/(LDi-HDo)] morph ratio is a response to the environmental conditions under which the plant is grown. If the ratio is > 1.00, production of (HDi-LDo) diaspore morphs is higher than that of (LDi-HDo) diaspore morphs, and if www.nature.com/scientificreports/ the ratio is < 1.00, production of (HDi-LDo) diaspore morphs is lower than that of (LDi-HDo) diaspore morphs. Thus, a decrease in the (HDi-LDo)/(LDi-HDo) morph ratio has been demonstrated for increasing soil water 18,21 , nutrient 15,36,37 , density 16,19 and herbivory 18 stress. Likewise, in the polymorphic diaspore species C. arenarius soil water, nutrient and density stress caused a decrease in proportion of dispersal unit morph f (HDi-LDo) with increase in intensity of all the three stresses, whereas proportion of dispersal unit morph a (LDi-HDo) increased with increase in intensity of all the three stresses. Moreover, proportion of dispersal unit morph c was the same as that of dispersal unit morph f (i.e. decreased) or a (i.e. increased) with increase in intensity of stress. Overall,  Table 3). In the heterodiasporous species Lappula duplicicarpa 19 , Atriplex aucheri 38 and Heterosperma pinnatum 26 , the dispersal unit morph number ratio can vary, depending on the environment conditions under which the mother plants are grown. This diaspore number ratio also varies in the amphicarpic sensu lato species Catananche lutea 22 , Emex spinosa 39 and Gymnarrhena micrantha 40 , the amphicarpic sensu stricto species Amphicarpum purshii 41,42 and Mimulus nasutus with aerial chasmogamous and cleistogamous fruits 43 . In amphicarpic species, aerial diaspores www.nature.com/scientificreports/ may not be produced if growth conditions are highly stressful; thus, the aerial diaspore/subterranean diaspore ratio can be zero 41 . In the grass Bromus unioloides, the aerial CH/CL ratio varied with photoperiod and soil moisture conditions of the mother plants and ranged from 0 to 100% 44 . In H. pinnatum, the proportion of achenes with awns increased with decrease in R 0 (fitness), i.e. increase in stress, and was nearly 1.0 in environments in which R 0 < 1.0, i.e. a negative correlation between fitness and proportion of awned achenes. For this species, the authors reported that (depending on stress level) individuals can produce any ratio combination of dimorphic   18 and Lappula duplicicarpa 19 . The effect of dispersal unit morph on CV of dispersal unit number and mass varied by treatment, with dispersal unit morph f having the highest CV for number (range 1.24 to 1.62) and mass (range 1.10 to 1.53) and dispersal unit morph a the lowest CV for number (range 0.25 to 0.41) and mass (range 0.49 to 0.76). The CV for dispersal unit morph c for number (range 1.04 to 1.26) and mass (range 0.94 to 1.18) was between those of dispersal unit morphs a and f. The effects of dispersal unit morph on the CV of dispersal units with a fruit varied by treatment in the same manner as that of the dispersal unit morphs, i.e. f > c > a.
The shift in proportion of heteromorphic diaspores is of ecological significance in the dispersal and germination stages of a plant's life history 2,46 . In stressful and nonstressful environmental conditions, dispersal unit morph a of C. arenarius is a low-risk phenotype that germinates in a proven site occupied by the mother plant, and dispersal unit morph f is a high-risk phenotype that explores new habitats away from the mother plant. However, as expected, both number and proportion of morph f were higher in nonstressful than stressful situations. These results are in agreement with those of Venable et al. who found that populations of H. pinnatum with a higher proportion of central achenes with awns (high-dispersal, low-dormancy achenes) occurred in sites where precipitation was relatively high 26 . Thus, more high-risk diaspores are available for colonization of sites Table 1. Summary of two-way ANCOVAs showing effects of different levels of water (W), nutrient availability (N), pure (D 1 ) or mixed (D 3 ) density, plants from the three dispersal unit morphs (M) and their interactions on proportion of each dispersal unit morph from offspring of Ceratocarpus arenarius. There was no significant effect of initial seedling size, which was used as a covariate, on any of the variables in the analyses (data not shown). www.nature.com/scientificreports/ away from the maternal plants of C. arenarius in nonstressful than in stressful conditions. Consequently, with an increase in favorable conditions for growth and survival of her offspring the mother plant devotes proportionally more resources to the high-risk morph. As such, then, production of dispersal unit morph a is a "timid" strategy for ensuring reproduction, while production of dispersal unit morph f is a "bold" strategy 47,48 . Moreover, production of dispersal unit morph c is a transitional type between "timid" and "bold" strategies. However, regardless of environmental conditions (stressful vs. nonstressful) presumably some degree of bet-hedging capability was retained: under no environmental condition was there a complete shift to production of only one morph. This retention of a (HDi-LDo)/(LDi-HDo) morph ratio greater than zero under very stressful environmental conditions also occurs in the heterodiasporous plants Atriplex sagittata 15,36 , Leontodon saxatilis 16 , Diptychocarpus strictus 18 and Lappula duplicicarpa 19 .
Variable diaspore functions in heteromorphic species usually are explained in terms of bet-hedging 2 , an adaptation to highly variable and unpredictable environments in time or space 14,49 . The success of C. arenarius in the cold desert of Central Asia might be, at least in part, attributed to production of a range of variation in dispersal unit morphs and the ability to shift the dispersal unit proportion under changing environmental conditions.

Materials and methods
Dispersal unit collection and field site description. Freshly-matured dispersal units with khaki-coloured persistent bracteoles (Fig. 1) (2) each of five aerial diaspore morphs in the gradient of dispersal units above the soil surface ( Fig. 1) 33 . Dispersal unit morphs a, c and f were bulked separately and stored in paper bags at room conditions (16-30 °C, 10-40% RH) until used in experiments.
The area of the Junggar Basin where the dispersal units were collected has a temperate continental climate. Mean annual temperature is 6.6 °C, and average annual precipitation (including rain and snow) is 160 mm, about two-thirds of which falls in spring and summer, and the snow that falls in winter begins to melt in March or April 50 . Annual potential evaporation is > 2000 mm 51,52 . Additionally, rainfall is highly variable among seasons and years, but generally, rainfall is higher in spring than in autumn 53,54 . Further, water from snowmelt increases water availability in spring. experimental design. Our observation in the experimental garden and in a diversity of habitat types in the field, levels of soil moisture, nutrients and competition (density) are the important contributors to the phenotypic responses (plant size and dispersal unit production) of C. arenarius plants.
To obtain seedlings from dispersal unit morphs a, c and f, on 4 April 2017, 1,600 isolated fruits (i.e. dispersal units with their bracteoles removed) of each dispersal unit morphs a, c and f were incubated in 0.5 mmol·L −1 GA 3 solutions at 25/15 °C in light/dark (12/12 h) for five days. Then, during 10-15 April 2017, 3,720 seedlings (10 seedlings per pot × three dispersal unit morphs × 124 pots) that germinated on the same day were transplanted into pots (20 cm deep and 24 cm in diameter) with drainage holes in the bottom. The pots were filled with a mixture of 70% grey desert soil and 30% desert sand; sand was added to improve soil aeration and drainage. Seedlings were exposed to near-natural temperatures in a metal framehouse with no heating or cooling with sides open, in the experimental garden on the campus of Xinjiang Agricultural University in Urümqi, which is located near the southern edge of the Junggar Basin 19,55 . Before the beginning of each treatment (see below), the soil was watered daily to field capacity. To prevent variation in initial seedling size, seedlings of the same size in each pot were kept (i.e. one plant per pot), and the others were removed from the pots and discarded, except in the density treatment with ≥ 1 to 9 seedlings per pot (see below) 18 . A factorial design was used in the soil water content, nutrient supply and density treatments, which were initiated when plants were in the four-leaf rosette stage (during 25-30 April 2017, 15 days after transplanting) 19 . In addition, the number of individuals surviving from the four-leaf rosette stage to reproduction was monitored. Fifty to 62.5% of seedlings from the three dispersal unit morphs at high levels of nutrient supply and ≥ 90% of seedlings from the other treatments survived and reproduced; the experiment ended on 3 October 2017.
Each treatment was an independent test. While one factor was manipulated, the others were kept at high or moderate levels to prevent limitation of plant growth. That is, the soil was kept moist except in the soil water content treatment, one plant per pot except in the density treatment and no nutrient supply added except in the nutrient supply treatment. Weeds were removed from the pots as needed to prevent competition from them 18,19 .
Soil water content. Four moisture levels were applied to the plants: watered to field capacity every day (high water supply, i.e. the lowest stress), every three days, every six days or every nine days (i.e. the highest stress) 18 . Evidence for water stress was shown by the smaller size of plants in the 9-day than 6-day and in the 6-day than in the 3-day watering treatment. www.nature.com/scientificreports/ shown by smaller size of plants in the 0 than in the 3 g nutrient addition and in the 3 g than in the 6 g nutrient addition.
Density. Density experiments were carried out using pure (D 1 ) and mixed (D 2 and D 3 ) stands. D 1 was used to study the effect of density stress on vegetative and reproductive dry mass of plants from the different dispersal unit morphs. Four levels of intramorph D 1 were used: (1) one plant in the pot (target plant) (i.e. the lowest stress), (2) one target plant and one other plant from the same morph, (3) target plant surrounded by three plants from the same morph and (4) target plant surrounded by seven plants from the same morph (the highest stress) (Fig. 6A). Plants from two dispersal unit morphs were combined in a 1:1 ratio at three total densities [two (i.e. the lowest stress), four and six (i.e. the highest stress) plants per pot]. Plants from the two dispersal units were planted alternately, and combinations of plants from any two of the three dispersal unit morphs were used: (1) (Fig. 6B).
Competition among plants from the three dispersal unit morphs (D 3 ) was tested in a 1:1:1 ratio at three total densities [three (i.e. the lowest stress), six and nine (i.e. the highest stress) plants per pot] (Fig. 6C).

Measurements
Each level of soil water content treatment was replicated eight times (eight plants were harvested to measure all indices at maturity), and nutrient supply treatment was replicated eight times at low and intermediate levels and four times at high levels. Except for eight replicates of one plant per pot in the D 1 treatment, the other levels of D 1 and each level of D 2 and of D 3 was replicated five times. We harvested each plant during 6-22 September 2017, when there was a total lack of chlorophyll (plant death). When all dispersal units were mature, plants in each treatment were excavated to a soil depth of 15 cm, and roots were gently washed free of soil 19 .
Dry mass accumulation and allocation. At harvest, plants in each treatment were collected separately and divided into root, stem, leaf and reproductive organs [i.e. including all dispersal unit morphs and their glochids (if present) and trichomes on bracts] and oven-dried to constant mass at 80 °C for 48 h in paper bags. Once dry, all parts were weighed using an electronic balance (0.0001 g). Vegetative mass was calculated as sum of mass of roots, stems and leaves. Total mass is vegetative mass plus mass of reproductive organs. Allocation to roots, stems, leaves and reproductive organs was calculated by dividing mass of each organ by total mass and then multiplying by 100 18 . Additionally, reproductive mass was expressed as coefficient of variation (CV = standard deviation/mean) in plants derived from each dispersal unit morph.
Dispersal unit and fruit production. Number. Number of dispersal unit morphs a, c and f on the plant in each replicate at different levels in the treatments was recorded. Total number of dispersal units per individual was determined by summing number of all dispersal unit morphs per individual. By randomly sampling and making observations, we found that if there is a fully developed fruit/seed in a dispersal unit the fruit is inflated and can be seen clearly through the bracteoles under a fluorescent light. However, if a fruit/seed is not present in a dispersal unit the bracteoles are not inflated. Thus, the number of each kind of dispersal unit morphs (with and without a fruit) on each plant was counted using the observation method described above. Percentage of each of the three dispersal unit morphs (with + without a fruit) (NP j ) and percentage of each of the three morphs with a fruit (NP j ʹ) was calculated as NP j = (ND j / NT) × 100 and NP j ʹ = (ND j ʹ/ NTʹ) × 100, respectively, where ND j is total number of each of the three dispersal unit morphs (with + without a fruit) and ND j ʹ number of each of the three dispersal unit morphs with a fruit, NT total number of the three dispersal unit morphs and NTʹ total number of the three dispersal unit morphs with a fruit per individual and j the given dispersal unit morph (a, c or f).
Mass. Each dispersal unit morph per individual was weighed using the electronic balance (0.0001 g). Mass percentage of each of the three dispersal unit morphs per individual was calculated by dividing mass of each dispersal unit morph by total mass of the three dispersal unit morphs and then multiplying by 100.
Additionally, number and mass of dispersal unit a, c and f morphs was expressed as CV (i.e. standard deviation/mean) in plants derived from each dispersal unit morph.
Statistical analysis. A one-way ANOVA was used to analyze differences in plant dry mass; allocation of dry mass to roots, stems, leaves and reproductive organs; number, proportion, mass and proportion of mass of each dispersal unit morph a, c and f; total number and mass of the three dispersal unit morphs; and number and proportion of each dispersal unit morph with a fruit per individual among levels in each treatment. Data were arcsine (percentage data) or log 10 (other data) transformed as needed before analysis to approximate a normal distribution and homogeneity of variance to fulfill assumptions of a one-way ANOVA. If variance of transformed data was still not homogenous, treatment differences in these variables were assessed using the Kruskal-Wallis non-parametric test. Tukey's HSD test was performed for multiple comparisons to determine significant differences among treatments 56 . Two-way ANCOVAs or three-way ANCOVAs were used to determine whether proportion of each dispersal unit morph differed among levels in each treatment. For "soil water content", "nutrient supply", "D 1 " and "D 3 " treatments in the two-way ANCOVAs, plants from the three dispersal unit morphs and treatment (soil water content, nutrient supply, D 1 and D 3 ) were considered fixed effects, and all interaction terms were included. For the "D 2 " treatment in the three-way ANCOVAs, plants from the three dispersal unit morphs, treatment (levels of D 2 ) and combination with any two morphs of the three dispersal unit morphs [D 2(a:c) , D 2(a:f) and D 2(c:f) ] were Scientific RepoRtS | (2020) 10:11142 | https://doi.org/10.1038/s41598-020-67380-0 www.nature.com/scientificreports/ considered fixed effects, and all interaction terms were included. Total mass was used as a covariate to minimize the effects of variation in plant size on the proportion in the analyses. Pearson correlations were used to determine the relationships between proportion of any two of the three dispersal unit morphs and between mass of total plant and proportion of each dispersal unit morph in each treatment (Supplementary Tables S1-S5). All data analyses were performed with the software SPSS 19.0 (SPSS Inc, Chicago, IL, U.S.A.). Values are means ± 1 s.e.