1. Introduction
The scarcity of water resources and high irradiance and the temperature during the summer are characteristics of different regions of the world, such as the Mediterranean area [
1] or California area [
2]. These areas are considered centers of agriculture in their respective countries and produce a very significant portion of the agricultural national production. It has been proved that effects of drought have a huge impact on the morphology, physiology and biochemistry of fruit trees [
3,
4,
5,
6]. Within the actual climate warming, drought periods in arid and semi-arid regions worldwide are predicted to occur more frequently [
1]. The
Prunus genus englobes important fruit trees (peaches, cherries, apricots, plums) and also important nut trees, such as almond, are cultivated worldwide. At the same time, these species represent the most economically important fruit of these regions, for example, almond generated USD 5.60 billion in cash receipts in California, [
7].
In the cultivated peach (
P. persica L.) and its relative species, a suite of morphological and physiological adaptations to drought have been described, allowing them to survive a water stress situation [
8,
9,
10,
11,
12]. The degree of adaptation to drought may vary considerably among and within species. In dry regions, such as the Mediterranean area, the choice of proper rootstocks with multiple tolerances to stresses is crucial to preventing future problems in orchards and to reducing management costs [
13,
14]. Due to this fact, classical breeding programs tried to develop new rootstocks as part of their breeding goals to improve resistant to drought [
15]. Suitable rootstocks could increase adaptation to drought, heavy soils, waterlogging, alkalinity, vigor control and soil fungal diseases [
16].
To study drought response, several parameters and their relations such as stomatal conductance (G
S), net photosynthetic rate (P
N), water distribution between symplast and apoplast, leaf turgor, transpiration rate (E), synthesis of abscisic acid, intercellular CO
2 concentration (Ci), electron transport rate, carboxylation efficiency, intrinsic water-use efficiency have been studied in fruit trees [
6,
17,
18,
19,
20]. Under drought stress conditions, all metabolic processes, including photosynthesis, are negatively affected [
20,
21]. In addition, Rieger and Duemmel [
8] concluded that shoot characteristics measured via carbon assimilation rate and leaf conductance were more closely associated with drought adaptation than root characteristics in six
Prunus species from divergent habitats. From all measured parameters, only a fraction of root biomass in fibrous roots was correlated with drought resistance. According to the variation in leaf characteristics, the authors assumed that genetic improvement of drought resistance of stone fruits may be achieved by incorporating xerophytic leaf characteristics into scion cultivars [
8].
Torrecillas et al. [
22] used the two almond (
P. dulcis) cultivars “Ramillete” and “Garrigues” to study water stress mechanism on almonds. The authors evaluated osmotic adjustment and turgor maintenance, leaf turgor and stomatal conductance. According to the results, different mechanisms to resist water stress seem to have been developed by these cultivars. These differences could be associated to differences in cell wall thickness or in cell wall structure [
23]. In addition, higher leaf conductance values for plants grafted onto “Garrigues” were observed by Alarcón et al. [
24] after a short and long time of water deficit exposure, in comparison with plants grafted onto “GF 677”. The authors suggested that “Garrigues” seedlings, as a rootstock, are less drought tolerant than “GF 677”, because the latter avoided the loss of water via transpiration and maintained the leaf water content under stress [
24].
Four Prunus rootstocks (“GF 677” (
P. persica ×
P. dulcis), “Cadaman” (
P. davidiana ×
P. persica), “ROOTPAC
® 20” (
P. cersífera ×
P. dulcis) and “ROOTPAC
® R” (
P. cersífera ×
P. dulcis)) were submitted to drought stress to study physiological, biochemical and molecular parameters and to improve water-use efficiency (WUE) in stone fruit [
4]. According to the results, a lower net photosynthesis rate, stomatal conductance and transpiration rate, and higher intrinsic leaf WUE (A
N/G
S) were observed in the stressed plants. In addition, these results showed that accumulation of sorbitol and raffinose, proline, and the increase in expression of the Δ
-1-pyrroline-carboxylate synthase (P5SC) gene could be used as markers of drought tolerance in peach cultivars grafted on
Prunus rootstocks [
4].
More recently, Bielsa et al. [
25] studied a collection of wild-relative species and cultivated hybrid rootstocks of
Prunus to estimate their water use efficiency (WUE) and genetic variation in two drought-induction genes. Results showed that almond and peach wild-relative species had the highest WUE in comparison with hybrid genotypes, with almond wild-relative species being the best candidates to develop new cultivars with drought tolerance. Moreover, few differences were observed in the promoter regions of the studied genes, which could be also used to improve
Prunus rootstock germplasm.
In general, a common pattern between these agronomic experiments studying response to drought (and other abiotic stresses) is that the collected data have been based on repeated measurements on the same plant (either across time or across space). It is often plausible that two measurements taken closer in time on the same unit (e.g., plant) are likely to be stronger correlated than measurements taken further apart in time. In this sense, an appropriate statistical methodology is essential for an efficient and reliable analysis of these agronomics experiments which are often expensive and time-consuming [
26]. However, studies about modelling the temporal physiological response of the most important
Prunus species under drought stress via mixed model are still scarce.
The purpose of the present study was to analyze several physiological parameters in different Prunus genotypes using a mixed model approach to determine if these Prunus rootstocks and wild species respond similarly to drought stress and recover afterwards or, contrarily, if different patterns of response can be observed. In addition, the study tries to answer the question of which parameters best describe the drought and recovery pattern.
2. Materials and Methods
2.1. Plant Material
The plant material used represents commercial rootstocks from the most important
Prunus species, including the apricot (
P. armeniaca L.) “Real Fino”, the peach cultivar “Montclar”, the hybrid peach × almond “GF 677” and the almond “Garrigues”. In addition, the wild Mediterranean almond species,
Prunus webbii [
27], was also included in the study. “GF 677” (Paramount
® in the USA) is a natural hybrid selected by INRA. It is a very vigorous rootstock (10–15%) more vigorous than peach seedlings [
28] with a well-developed root system that ensures good anchorage. “Montclar
®” was selected at INRA in 1960. It has high seed production and increased vigor in scion cultivars [
29]. It also exhibits great resistance to iron-induced chlorosis and has a good uptake of iron and magnesium from the soil. “Real Fino” is a traditional apricot seedling rootstock. It is specially adapted to light soils containing limestone due to its resistance to iron chlorosis [
30]. “Garrigues” is the most common almond rootstock in Spain. Seedlings from this cultivar are uniform with a strong, deep, root system. Finally,
P. webbii, is a wild almond species grown in a wide range of countries, from Balkan peninsula to Italy or Spain [
31]. This small and very thorny bush species with small leaves has been used as a source of new genes for improving important traits, such as self-compatibility or tolerance to adverse environment in almond [
32]. All plants from “Real Fino”, “GF 677”, “Garrigues”, “Montclar” and
P. webbii, were open pollinated seedlings collected from each individual mother tree located in the germplasm collection of the experimental field of CEBAS-CSIC at Santomera and Cieza (Murcia, Spain). Seedlings were germinated and planted in individual pots at the same time. At the beginning of the experiment, the development of the plants was similar across individuals. Note that we used the terms, species, cultivars and hybrid for different genotypes till now, as they were partly species while others were genotypes. To simplify presentation, we used the term “genotype” globally to describe differences between the five genotypes used.
2.2. Experimental Design
The experiment was performed in controlled greenhouse conditions in the experimental field of CEBAS-CSIC at Santomera (Murcia, Spain). Hourly means of temperature and relative humidity in greenhouse conditions were recorded to assure constant conditions during the experiment (
Figure S1). The greenhouse experiment was conducted to test the influence of three watering treatments (described below) on the above-described material. A total of nine tables were used within the greenhouse. Within a table, five zones were defined. Genotypes were randomly allocated to these zones. Within a zone, there existed eight pots planted with the same genotype and with one plant per pot. Thus, a total of 360 pots and plants existed. Pot size was 15 × 15 × 19.5 cm (3.5 L). Watering treatments were randomly allocated to tables (three tables for each treatment) according to a randomized complete block design. Furthermore, genotypes were randomly allocated to zones according to a randomized complete block design. The complete experimental design can be denoted as split-plot design with eight plants per zone.
At each time point, a random sample of plants per zone were measured. On each plant selected at one time point all measures described below were performed, these measures resulted in a loss of eight leaves in each sampled plant. Note that a loss of leaves each second day can result in damaged plants in the long term. Therefore, sampled plants varied between time points to avoid or reduce effects of repeated destructive measures (destructive effect) on the same plant. Further to note is that this is the main reason for planting more plants per genotype than plants sampled per time point. In addition, natural leaf drop-off should be expected as a result of management and the different treatments. To conclude, the experimental design can be denoted as split-plot design with repeated measures within a zone and across time.
2.3. Watering Treatments
The three different watering treatments included drought, recovery and well-watered (control). All plants were irrigated to full water capacity at the beginning of the experiment at day 0. Afterwards, the experiment and measurements started, as has been commented below. In the well-watered control treatment, plants were irrigated daily to full water capacity. The recovery treatment consists of exposing the plants to a period of drought until an established day and after that the plants are watered again (to full water capacity). The drought treatment consists of not watering plants at all. The established day of the watering day in the recovery treatment was selected according to the previous observed tolerance to drought of these genotypes, which is described below.
2.4. Time Points and Duration of Experiment
Measurements of the physiological parameters (described below) were taken repeatedly every second day in the selected plants, between 9 and 11 am (GMT). The duration of the experiment was 21 days for “Garrigues” and P. webbii and 17 days for the rest of the genotypes. The re-watering point was established at day 17 of the experiment for “Garrigues” and P. webbii (because of their drought tolerance) and day 13 in the case of the rest of the genotype. To facilitate interpretation of data, each time point was denoted with the word “Time” and the number of measures. As per example, day 1 (first measurement) is Time1, day 13 is Time7, and day 21 of the experiment is Time11.
At Time8 the LI-COR machine used during all the experiments was not available to the authors. To avoid loss of data, authors decided to use a different one for this day. As the machine effect cannot be separated from time point effects here (the second machine is used for time point 8 only), for these traits (net photosynthesis rate (PN), stomatal conductance (GS), and transpiration rate (E)) data from Time8 were dropped prior to analysis.
2.5. Physiological Parameters Measurement
A destructive (leaf water potential (ψw)) and five non-destructive (net photosynthesis rate (PN)), stomatal conductance (GS), and transpiration rate (E), chlorophyll content (ChC) and maximal photochemical efficiency of photosystem II (Fv/Fm) traits were conducted in this experiment and have been described in detail below.
The assessment of leaf ChC was realized on one fully developed leaf of the selected plants by means of a portable device SPAD-502 (Minolta, Kyoto, Japan). The SPAD-502 m is used extensively in research and agricultural settings as a rapid, inexpensive, and non-destructive method [
33]. The SPAD-502 m consists of two light-emitting diodes (LEDs) and a silicon photodiode receptor. It measures leaf transmittance in the red region (650 nm) and infrared region (940 nm) of the electromagnetic spectrum. A relative SPAD-502 m value (ranging from 0 to 99) is derived from the transmittance values, which is proportional to the chlorophyll content in the sample [
34,
35]. Three measurements were taken in different areas of the leaf, and the average value was obtained.
The chlorophyll fluorescence was analyzed on a second leaf by means of a portable Chlorophyll Fluorometer (Opti-Sciences, Hudson, NH, USA). This equipment obtains Fv/Fm = (Fm − F0)/Fm and thus is the ratio between the variable fluorescence (Fm − F0) and the maximal fluorescence. Fm is the maximal fluorescence intensity in a leaf adapted to darkness during 30 min, induced by a far red light excitation source (3000 μmol m−2 s−1) during 0.8 s. F0 is the minimal fluorescence intensity due to the exposition of a leaf to an actinic light source (400 μmol m−2 s−1).
The measurements of P
N, G
S, and E were all taken from the third leaf. The fourth unfolded leaf was selected and a portable LI-COR meter (LI-6400XT) was used following the method of Haider et al. [
36].
Finally, another leaf was used to measure ψ
w using a pressure chamber, following the recommendations of Turner [
37] to prevent leaf water loss during measurements. Leaves were selected at random from the middle third of the shoots. Note that due to the special leaf morphology of
Prunus webbii, only SPAD and F
v/F
m measures were evaluated for this genotype. The remaining (four) sampled leaves were stored for molecular analysis (data not used in the current study).
2.6. Modelling Analysis
Each trait was evaluated across the whole study period (from day 1 to day 21). The analysis was based on a mixed model approach using the following model:
where
,
, and
are the vectors of fixed effects of the
kth replicate for each time point, the vector of random effects of the
jth main-plot error within the
kth replicate (which corresponds to table effects for each time point) and the
ith zone within table
kj for each time point. Note that these are vectors with an effect for each time point.
is the vector of residual errors of the observation vector
. Note that residual error effects are the confounded effects of plot error and the plant effects.
is the vector of fixed effect of the
ith genotype,
is the vector of effect of the
jth watering treatment, and
is the vector of interaction effects between the
ith cultivar and the
jth watering treatment. Each vector includes effects for up to eleven time points. As time points cannot be randomized, a first order autoregressive variance–covariance structure was fitted to all random effects. This accounts for correlations due to the repeated measures structure of the data. For the residual error effects, a homogeneous and heterogeneous first order autoregressive variance–covariance structure was fitted. The latter means that the model allows for time-point specific variances. The model with the smaller Akaike information criterion (AIC) [
38,
39] was chosen. Residuals were checked graphically for deviations from normal distribution and homogeneous (or the fitted heterogeneous) variance. For the traits E and G
S, data have been logarithmical transformed prior to analysis. For the trait Fv/Fm, data were logit transformed prior to analysis. Means were back transformed for presentation purpose only. Back-transformed values represent the median of the distribution on the original scale. Standard errors were back transformed using the delta method. A letter display based on Fisher’s Least Significant Difference (LSD) multiple comparison tests were used in case of finding significant effects via F-test. Note that data were unbalanced at time point 10 and 11. Thus, genotype-by-treatment means across time were not estimable. Therefore, means across the first nine time points were estimated. Additionally, specific contrasts were estimated to account for the fact that drought took four days longer for drought-tolerant genotypes. All analyses were performed using the procedure PROC MIXED within SAS [
40].