Introduction

Six Mediterranean countries, Spain, Turkey, Egypt, Morocco, Italy and Israel, were among the major producers for mandarins, sweet oranges, grapefruits, lemons and/or limes in the world in 2012 (FAOSTAT 2015). However, this region is affected by water scarcity what forces the use of marginal water for irrigation (Oron et al. 2002) making it more vulnerable to salinization. In the future, this situation is expected to get worse in the Mediterranean region due to the effects of climate change on precipitation, evaporation, runoff and soil moisture storage (Paranychianakis and Chartzoulakis 2005). Citrus is classified among the most salt-sensitive tree crops (Maas 1993). Tree growth and fruit yield of citrus species are impaired at a soil salinity of approximately 2 dS m−1 soil saturation, without the concomitant expression of leaf symptoms (Bingham et al. 1974; Cerdá et al. 1990). Therefore, the adoption of both intense management practices and the use of salt-tolerant genotypes are generally recommended to maintain citrus productivity and to ensure land sustainability.

Salt tolerance in a crop species is conceived as the ability to reduce yield losses when exposed to salinity. Since citrus cultivars are always vegetatively propagated by bud grafting on a seedling rootstock, salt tolerance is usually one of the target traits in rootstock breeding programs taking advantage of the wide genetic diversity and adaptations of Citrus spp. and citrus relatives (Herrero et al. 1996). Thus, Cooper et al. (1956) early classified citrus species used as rootstocks in three groups: Citrus reshni (Cleopatra mandarin) as good salt tolerant, Citrus volkameriana (Volkamer lemon) and Citrus aurantium (sour orange) as medium salt tolerant and Poncirus trifoliata (trifoliate orange) as poor salt tolerant. However, salt tolerance in citrus can display large variation from one area to another (Maas 1993) probably the cause of direct and indirect interactions between salinity and other physical abiotic stresses like poor soil drainage, drought, irradiance, leaf temperature and atmospheric evaporative demand (Syversten and Levy 2005). Moreover, genetic diversity for salt tolerance among accessions may also exist within each species, as it was recently found in the “salt-sensitive” P. trifoliata regarding Cl− and Na+ excluding capacities (Sykes 2011). From the agronomic point of view, salt tolerance should be considered in terms of fruit yield; however, this type of evaluation is extremely lengthy and difficult in citrus progenies (Raga et al. 2014, García et al. 2002). Therefore, it is necessary to approach the genetic analysis of salt tolerance in non-grafted progenies derived from salt-tolerant citrus rootstocks where it is reasonable to expect segregation for genes controlling Cl− and Na+ homeostasis, vegetative growth, water potential and other traits possibly related to the salt tolerance conferred by the citrus rootstock. Since these traits are quantitative, citrus rootstock breeding could benefit from their quantitative trait locus (QTL) analyses. In fact, QTL analysis of salt tolerance is considered a prerequisite to allow cost-effective applications of genomic-based approaches to breeding programs (Collins et al. 2008) which are particularly long and expensive in the case of citrus rootstocks. The detection of genomic regions containing QTLs controlling salt tolerance would be useful in the search for candidate genes and the implementation of marker-assisted selection schemes in breeding programs of citrus rootstock. An early attempt of QTL analysis of salt tolerance was reported by Tozlu et al. (1999a, 1999b) using a clonally propagated BC1 progeny derived from a citrus rootstock (from P. trifoliata) and a citrus variety (from Citrus grandis). However, C. grandis shows no apomictic reproduction and citrus rootstocks must be able to reproduce asexually through seed by apomixis (adventitious embryony from nucellar cells) to ensure the cheap establishment of well-rooted, uniform orchards. Therefore, a citrus progeny derived from two citrus rootstocks, being at least one of them salt tolerant, would be a more realistic approach. Besides, genotype replication through seed would be possible here by selecting apomictic hybrids within the progeny for the genetic analysis. We have chosen this approach by using a segregating population derived from two citrus rootstocks: P. trifoliata and C. reshni (the well-known salt-tolerant rootstock Cleopatra mandarin). However, two problems had to be faced in the experimental design: (1) the assessment of nucellar seedlings and (2) the limited number of apomictic hybrids in the progeny. The former can be solved by marker analysis (Ruiz et al. 2000), but the latter, which greatly affects the power of detection of QTLs (Asins 2002), is limited by the size of the original population and the accompanying, maintaining costs. To counteract this limitation, we used replicated progenies which can bring about a major reduction in the number of genotypes that need to be scored, and we carried out two different salt tolerance experiments to test the repeatability of QTLs detected for leaf weight traits.

The objective of this study is the genetic analysis of salt tolerance in terms of vegetative and physiological traits in a progeny derived from two well-known citrus rootstocks, Cleopatra mandarin and trifoliate orange. Both show apomictic reproduction but are genetically distant and differ in many morphological, physiological and agronomic traits, including salt tolerance.

Materials and methods

A reference population that consists of 151 hybrids (RxPr) which had been previously genotyped (Raga et al. 2012) was used for two salt tolerance experiments. This population was obtained at IVIA (Valencia, Spain) by controlled crosses between Citrus reshni Hort. ex. Tan. (Cleopatra mandarin) as female (salt tolerant and apomictic) parent and two apomictic and disease-resistant varieties of P. trifoliata (L.) Raf. (trifoliate orange): Flying Dragon (83 hybrids) and Rich (68 hybrids) as pollinators.

Plants were cultured into pots (3.5 L), and repetitions (nucellar plants) of each genotype were randomly selected to establish two treatments: control (no NaCl added) and saline irrigation (25 or 30 mM NaCl, depending on the experiment). Zygotic seedlings were identified by analysis of molecular markers (Ruiz et al. 2000) and discarded to keep nucellar seedling only.

The first experiment (Exp. 1) was run under a screenhouse during 3 years (2006–2008). It was started with 1-year-old seedlings from 24 apomictic hybrids in 2006. To increase the number of apomictic genotypes in the progeny that was limited by the juvenility and alternate bearing of the reference population, 1-year-old seedlings from new apomictic RxPr hybrids were added to this experiment in 2007 (18) and in 2008 (28). Thus, there was a total of 533 plants in experiment 1 during 2008 that corresponded to the nucellar progeny from 70 RxPr hybrids and two parents, Cleopatra mandarin and Flying dragon, as controls. Plants were grown with a sterilized substrate mix (50 % peat, 30 % coconut fibre, 15 % sand and 5 % perlite) and placed following a completely randomized design with three to five repetitions per hybrid and treatment. Salt treatment (25 mM NaCl) was applied through irrigation between June and September. Plants were irrigated with 250 ml each of just tap water (0.1 dS/m; pH 6.0, as control) or salted water (salt treatment) three times a week during the salinity treatment. NaCl concentration was increased gradually from 10 to 25 mM (3.2 dS/m; pH 7.5) in a week (conductivity meter HANNA HI9033 multi-range instruments). Both control and salinity-treated plants received the same pest, disease and weed control; fertilization; and pruning work. Fertilizer was proportioned automatically by mixing in a 1:100 proportion the stock solution A (200 mM NH4H2PO4) and solution B. The stock solution B contained 12 M Ca(NO3)2, 2.6 M KNO3, Sequestrene 138Fe [36 g L−1; (Syngenta, Madrid, Spain)] and 5 % of microelement solution. This solution contained 0.3 mM Cu SO4·5H2O, 3.1 mM Zn SO4·7H2O, 109 mM Mn Cl2·4H2O, 92 mM BO3H3, 2 mM NH4MoO4 and 0.4 mM V2O5. Salt-treated plants were grown similarly to control plants from October 2006 to June 2007 and from October 2007 to June 2008. The following vegetative and physiological characters were evaluated (Table 1): dried and fresh leaf weight in grams (DLW and FLW, respectively); accumulated dry matter percentage (%DMA) measured as the difference between dry leaf weight at the beginning and at the end of salt treatment; leaf water content (LWC) evaluated as the difference between leaf fresh and dried weights; leaf area (LA) in square centimetre and the increment in LA during the saline treatment (LAG) in cm2; and leaf colour parameters LCL*, LCa* and LCb* defined by Hunter (L* a* b*) coordinates arranged in a Cartesian system and Cr* and Hue* defined by a cylindrical coordinate system (L* C* h), where Cr* (Chroma) represents colour intensity (0 to 60) {Cr* = √[(a*)2 + (b*)2]} and Hue* (Hue angle) represents leaf colour (0° to 360°) [Hue* = tan−1 (b*/a*)] (HunterLab 1996); leaf Cl− concentration (mg/L) was measured as described by Gilliam (1971) using a Sherwood chloride analyzer 926.

Table 1 Correspondence between the full trait names and their abbreviations
Full size table

The second experiment (Exp. 2) was run under a greenhouse with controlled humidity and temperature during 3.5 months. Plants were distributed in six blocks following a split-plot design (one repetition per block and treatment) using salinity treatment as the main plot. Seventeen-month-old, nucellar seedlings from 41 apomictic RxPr hybrids were grown in substrate until the experiment was started. Then, substrate was changed to sand and the salt treatment (half diluted Hoagland solution plus 25 mM NaCl) started on June 28, 2010, and reached 30 mM NaCl on July 5, 2010. Plants were harvested for evaluation at the end of the experiment (October 14, 2010).

Several vegetative and physiological traits were evaluated in Exp. 2 (Table 1): defoliation index (DI, visually assigned from 0 to 10), dried and fresh leaf weight in grams (DLW and FLW, respectively), plant height (H) in centimetre, stem diameter (SD) in millimetre and leaf water potential (Ψ) in megapascal measured around solar noon using a Scholander pressure chamber (Scholander et al. 1965) and following the precautions described in Turner (1981). Leaf water potential was measured 7 and 80 days after salinity treatment imposition (Ψ7 and Ψ80, respectively). The difference between both values Ψ 7 and Ψ 80 was named d Ψ. Tissue samples of leaf and root were fresh weight determined, oven dried for 48 h at 80 °C, weighted (dry weight) and prepared for mineral analysis by digestion in a HNO3/HClO4 (2:1, v/v) solution. Inorganic solutes (Al, B, Ca, Cr, Cu, Fe, K, Li, Mg, Mn, Mo, Na, Ni, P, S, Sr, Ti, V, Zn) were determined in parts per million by inductively coupled plasma (ICP) spectrometry (Ionomic Service; CEBAS–CSIC, Murcia, Spain) in leaf (L) and root (R) tissues. The differences between leaf and root concentrations (L-R) of some elements (Na, K, Ca, P) and the fresh and dried weights of the root system (total fresh root weight (TFRW) and total dried root weight (TDRW)) were also considered; leaf and root water contents were estimated as the difference between fresh and dried weights (LWC and root water content (RWC), respectively).

For comparative purposes, three traits were evaluated in both experiments as salt tolerance indicators: FLW, DLW and LWC. Other traits are specific of experiment 1 or 2 (Table 1).

For the first experiment, fixed effects for genotype (G), treatment (E) and year (Y) and all their possible interactions, i.e. GxE (genotype by treatment), GxY (genotype by year), ExY (treatment per year) and GxExY (genotype by treatment and year), were analyzed by a repeated measures approach using trees (within each genotype and treatment) as random subject factor and first-order autoregressive covariance structure between measurements taken from the same tree over the years. Pearson correlation coefficients and principal component analysis based on the correlation matrix for the adjusted means were used to study the relations between the different traits. The statistical analysis of the second experiment followed its experimental design; i.e. blocks were random, and to study the GxE interaction, the effects of genotype and treatment were classed as fixed. Considering RxPr hybrid genotypes as a random effects factor, broad-sense heritability (H 2) was estimated for all traits (those evaluated in 2008 in experiment 1 and all evaluated in experiment 2) for nucellar seedlings (repetitions) derived from apomictic RxPr hybrids under control (H 2c) or salinity (H 2s) conditions, based on the genotypic (V G) and environmental (V E) variance estimators calculated by minimum variance quadratic unbiased estimator (MIVQUE), as previously reported (Villalta et al. 2007).

QTL analyses were carried out using genotypic and map data from Raga et al. (2012) and adjusted means of traits by interval mapping (IM), multiple QTL mapping (MQM) and Kruskal-Wallis (KW) procedures in MapQTL® 6 (Van Ooijen 2009). Following the strategy explained by Asins et al. (2015), two different QTL detection procedures were used. First, we analyzed the data for each parental meiosis separately i.e. a “two-way pseudo-testcross” analysis (Grattapaglia and Sederoff 1994) what provides the computation advantages of the two genotype QTL models, but the disadvantage of losing power because intralocus interaction is ignored (Van Ooijen 2009). And second, to consider intralocus interaction and unlinked markers and to avoid computation difficulties derived from abundance of dominant markers and segregation distortion, the KW QTL mapping procedure was also used providing complementary genetic information on the highly significant genotypic means. For IM and MQM, a 5 % experimentwise significance level was assessed by permutation tests. These logarithm of the odds (LOD) critical values ranged from 1.2 to 2.5 depending on the trait and linkage group. MQM was used whenever IM detected more than a QTL in the linkage group under study. Only QTLs with LOD ≥2.2 are reported here. For the KW procedure, a cross pollination design of marker genotypes was used and the significance level was fixed at the stringent value of p ≤ 0.005. Both procedures used different map files accordingly to their respective locus files (Raga et al. 2012). Cleopatra maps contained 81 or 80 markers, distributed along 9 or 12 linkage groups, covering 920.390 or 793.892 cM of the genome depending on the pseudo-testcross or cross pollinated genotype coding, respectively. Similarly, Poncirus maps contained 72 or 73 markers, distributed along 11 or 12 linkage groups, covering 682.627 or 729.883 cM of the genome.

Results

Tables 2 and 3 summarise the results from the mixed model analysis. As expected, year effects were significant for all traits in Exp. 1 (Table 2). Genotype effects were significant in most traits. The only exceptions were dΨ, L_Al, R_Al, R_Cu and R_Zn. A smaller proportion of traits showed significant treatment (E) and GxE effects in the Exp. 2 than in Exp. 1. In fact, common traits fresh and dried leaf weight (FLW, DLW) and leaf water content (LWC) showed significant E and GXE effects only in the longest experiment (Exp. 1); however, heritability estimates (also in Tables 2 and 3) were similar between experiments, particularly under salinity. The only vegetative traits significantly affected by salinity in Exp. 2 were those involving the root growth (TFRW, TDRW, RWC). Comparing leaf traits in both experiments, trifoliate parent means for FLW, LWC and DLW were always smaller than those of Cleopatra parent under both treatments indicating the existence of clear developmental and vigour differences for the aerial part of the plant between both rootstocks at origin (Fig. 1). The control and salinity distributions of most relevant traits (TFRW, TDRW, RWC, Ψ7, L_Na, R_Na, L_B, R_B, L_Fe, L_K, R_K and Cl) are shown in Fig. 2. Salinity greatly changed the distribution of Cl, L_Na, R_Na, L_B, L_Fe and L_K. Under salinity, the trifoliate parent showed higher means for Ψ7, R_Na and Na+ and Cl− leaf concentrations than the salt-tolerant parent (Cleopatra) whereas under control, the lowest Na+ concentration corresponded to the trifoliate parent and the lowest Cl− concentration to Cleopatra. A transgressive segregation was clearly observed for root vegetative traits (TFRW, TDRW, RWC) under both conditions.

Table 2 P values for the significant effects in the mixed model analysis and heritability estimates (H2_C and H2_S, for control and salinity treatments, respectively) of evaluated traits in experiment 1
Full size table
Table 3 P values for the significant effects in the mixed model analysis and heritability estimates (H2_C and H2_S, for control and salinity treatments, respectively) of evaluated traits in experiment 2
Full size table
Fig. 1
figure 1

Distributions of traits studied in both experiments 1 and 2 depending on the treatment (control, or salinity, indicated by white or grey bars, respectively). Fresh leaf weight, leaf water content and dried leaf weight of experiment 1 correspond to a–c, respectively, while those for traits in experiment 2 are depicted in d–f, respectively. The means for the parents: trifoliate orange (P) and Cleopatra mandarin (C) are indicated for control and salinity (in italics). Absolute frequencies of each phenotypic class are indicated at the Y axis

Full size image
Fig. 2
figure 2

Distributions of most relevant traits depending on the treatment (control or salinity, indicated by white or grey bars, respectively): leaf water potential after 7 days of treatment (Ѱ_7, a), total fresh root weight (TFRW, b), leaf Cl concentration (Cl, c), total dried root weight (TDRW, d), leaf Na concentration (L_Na, e), root water content (RWC, f), root Na concentration (R_Na, g), leaf B concentration (L_B, h), leaf K concentration (L_K, i), leaf Fe concentration (L_Fe, j), root K concentration (R_K, k) and root B concentration (R_B, l). The parent means, trifoliate orange (P) and Cleopatra mandarin (C) are indicated for control and salinity (in italics). Absolute frequencies of each phenotypic class are indicated at the Y axis

Full size image

Correlation between the adjusted means of all traits was investigated within each treatment, and 99 pairs of traits were significantly correlated under both control and salinity conditions (Online Resource 1). Common traits between experiments were significantly correlated under both control and salinity conditions except for LWC that was only significantly correlated under salinity. When Pearson coefficients of these 99 pairs of traits were graphically represented (Fig. 3), only a few pairs involving Cl−, R_Na and L_K deviated from the diagonal meaning a change of relationship depending on the controls or salinity condition. Five of these trait pairs, involving leaf Cl− concentration, changed their sign when comparing control and salinity conditions. Principal component analysis of traits in Exp. 2 (greenhouse experiment) allowed the graphical representation of trait relationships (Fig. 4). Comparing both treatments, the positions of Ψ7, Ψ80, L_B, L_Na and L_Cu remarkably changed. Thus, correlation analysis (Online Resource 2) showed that the change of leaf water potential under salinity (dΨ_S) was mostly related (r = 0.9) to the first evaluation of leaf water potential (Ψ7_S) which appeared highly, negatively associated with the root concentration of B, Mo and Mn (also in Fig. 4). Under control, both measures of leaf water potential were significantly correlated (r = 0.56), and again, dΨ_C was closer to Ψ7_C than to Ψ80_C. Differently from salinity, Ψ7_C was related to vigour traits (height and stem diameter), and the closest nutrient was Na+ (L-R_Na_C, r = −0.53, and R_Na_C, r = 0.51). Noteworthily, this trait showed a high correlation to TDRW under salinity (r = 0.75), i.e. the larger the root, the higher the root [Na+]. LWC was not directly related to leaf water potential but to its difference between control and salinity, Ψ 80_C–Ψ 80_S (p = 0.008, r = 0.44 for LWC2_C and p = 0.029, r = 0.36 for LWC2_S).

Fig. 3
figure 3

Dispersion of Pearson’s coefficients for significantly (p < 0.05) correlated traits under both control and salinity conditions (trait abbreviations in Table 1)

Full size image
Fig. 4
figure 4

Graphical representation of principal component (PC) analysis of traits evaluated in experiment 2 under control and salinity (trait abbreviations in Table 1)

Full size image

A total of 98 QTLs were detected by IM and MQM procedures using the two-way pseudo-testcross codification of genotypes (Table 4) and 135 marker-trait associations by KW procedure using the four-class codification for codominant markers segregating for different alleles at both parents (Online Resource 3). The closest markers to 45 QTLs by IM corresponded to the most significant marker by KW (markers in italics in Table 4). Most of them were located in linkage groups 3 (from Cleopatra) and 7 (from both parents). The position and one LOD QTL interval of the main clusters of QTLs are shown in Fig. 5. To better visualize complementary genetic information from both QTL detection strategies, integrated C. reshni and P. trifoliata maps of linkage groups 10+5b, 7, 3 and 4c showing the position of QTLs detected by KW are also shown in Fig. 5. The traits studied in both experiments 1 and 2 (FLW, DLW and LWC) showed common QTLs, remarking their repeatability (Table 4, Online Resource 3 and Fig. 5).

Table 4 List of the positions (in cM) and nearest markers (locus) to QTLs detected by IM and MQM in the genetic linkage maps of parents (GM): trifoliate orange (p) or Cleopatra mandarin (r) previously reported by Raga et al. (2012)
Full size table
Fig. 5
figure 5

Distribution of significant QTLs (IM and MQM) from Table 3 along Poncirus (p) and Cleopatra mandarin (r) linkage groups. Each QTL is coded as the trait (trait abbreviations in Table 1), and one LOD interval is depicted below. The corresponding Kruskal-Wallis trait-maker associations (Online Resource 3) are displayed in the integrated p+r (Int.) linkage groups 10+5b, 7, 3 and 4c, between the maps of both parents

Full size image

The effects of Cleopatra allele substitutions at CMS30,160 and C2iC1i,470 (significant markers within 76.263–89.566 cM of linkage group r3) on significantly associated traits are clear. The same QTL allele at C2iC1i,470 increasing LA under salinity is also increasing FLW, DLW and LWC, in both experiments and under both treatments (Online Resource 3). Therefore, the same QTL allele was found associated with the highest mean for the traits related to plant vigour. In the case of the QTL at CMS30,160, the increasing allele for DLW2 was also associated with a higher defoliation index (DI) and a lower leaf boron concentration. For leaf water potential, only one QTL was detected at CR17,300 in linkage group 7 (Online Resource 3) where the genotype bd showing the lowest Ψ 80 under salinity (Fig. 6) also showed the lowest defoliation index (DI_S), the largest leaf area (LA_S, at p ≤ 0.05) and the highest leaf [Na+] (L_Na_S) suggesting the presence of a putative Na+ tissue tolerance QTL. Two QTLs were detected for dΨ in other linkage groups of the Poncirus map: in p4b (at Py65C, a Citrus tristeza virus resistance marker derived from a resistance analogue gene) and p4c (MIR164, a marker derived from the miR164 gene).

Fig. 6
figure 6

Comparison of genotypic means at CR17,300 in linkage group 7 for the following traits LAG_C (leaf area growth under control, from Exp.1) and LA_C (leaf area under control), LA_S (leaf area under salinity), L_Na_S (Leaf [Na+] under salinity), DI_S (defoliation index under salinity) and Ψ80_S (leaf water potential after 80 days of salinity treatment) from Exp. 2. *LA_S QTL at marker CR17,300 was at p ≤ 0.05

Full size image

Discussion

Peculiarities of citrus rootstocks and salt tolerance

In spite of the economic and social importance of salt tolerance for the citriculture, its genetic analysis has been scarcely approached in rootstock segregant populations. These populations commonly derive from intergeneric crosses between genotypes from Citrus spp. and P. trifoliata (a main source of disease resistances). Such wide crosses are advantageous and usual in rootstock breeding programs of horticultural crops (King et al. 2010). However, contrary to other horticultural crops where parents are extensively homozygous, citrus parents have a certain degree of heterozygosis giving rise to up to four segregating alleles per locus in the first generation what makes genetic analysis possible but prevents the estimation of additive gene effect. Other difficulties of genetic studies in citrus rootstock populations come from their partial apomicitic reproduction. The yield of nucellar seedlings through apomixis makes uniform citrus orchards possible, easy and cheap. Therefore, hybrids with no apomixis or low yield of nucellar seedlings are generally discarded for further agronomic evaluation, drastically reducing the size of the target population. Given that nucellar reproduction is not at random, but genetically determined (García et al. 1999; Nakano et al. 2008; Raga et al. 2012), some genotypes are observed in a lower frequency than expected causing segregation distortion in some genomic areas. There are other genetic factors responsible for segregation distortion in citrus populations (Bernet et al. 2010; Raga et al. 2012), all of them make QTL analysis less powerful. Therefore, to obtain more robust conclusions, two salt tolerance experiments using replications (nucellar seedlings) from mature, apomictic hybrids of the same population were run and compared here. Salt tolerance was evaluated in terms of vegetative traits, and for comparison purposes, FLW, LWC and DLW were measured in both experiments (Fig. 1) showing that both parents were clearly different, with or without salinity, for the aerial part of the plant; i.e. no salt tolerance behaviour can be deduced for any particular parent. On the contrary, in Exp. 2 where leaf and root weight traits were evaluated under the same environmental conditions, the trifoliate parent showed a slightly sensitive behaviour in comparison to the Cleopatra root weight changes (Fig. 2). In fact, FLW2, LWC2 and DLW2 were not significantly affected by salinity in this experiment while TFRW, TDRW and RWC were indeed (Table 3). Besides, salinity effects on ionomic and physiological traits such as Cl, L_Na, R_Na, L_K, L_B, L_Fe and Ψ 7 may further explain Cleopatra salt tolerance in terms of trait mean changes which were smaller than those for the trifoliate orange (Fig. 2). Therefore, the joint analysis of these traits in the progeny will help us to understand the multiple effects of salinity on the citrus plants beyond the comparison of the parental genotypes.

Salinity effects on trait distributions and relationships

Salinity changed the relationships between (Fig. 3) and among (Fig. 4) traits. Thus, Fig. 3 shows how Cl− association with leaf colour parameters LCb, LCL and Hue or with LWC and fresh leaf weight changed of sign when comparing control and salinity, pointing out the importance of Cl− homeostasis in the plant under both conditions and not only under salinity. Salinity strengthened the relationship between L_K and DLW, while decreasing that between R_Na and L-R_Na (Fig. 3); i.e. the Na+ gradient between roots and leaves was not so closely related to R_Na under salinity than under control, suggesting that some rootstock genotypes can keep larger [Na+] differences between them than others (i.e. significant GxE interaction for L-R_Na). In fact, results from experiment 2 (Table 3) showed significant GxE interaction for L-R_Na and other traits (DI, L_K, L_Li, L_Na, L-R_K, R_S, RWC, TDRW and TFRW). TDRW appeared as a main trait related to plant vigour (DLW, H, SD) under both control and salinity (salt tolerance) and the root Na+ concentration, or the gradient of Na+ concentration between leaf and root (L-R_Na+) could play a relevant role in that relationship as suggested by the correlation analysis (Online Resource 2). The ability of the rootstock to retain the toxic ions in roots as a mechanism of salt tolerance had been previously reported in citrus (Kostopoulou et al. 2014; Balal et al. 2012). On the other hand, the distributions of L_K (and L_Fe) were also clearly displaced towards larger values under salinity (Fig. 2). Salinity stress effect on the K+ concentration of the aerial part of the citrus plant was previously observed by several authors (Morinaga and Sykes 2001; Balal et al. 2012; Khoshbakht et al. 2015). The only leaf nutrient directly related to leaf dried weight under salinity was Fe, and the others (K, P and B) were indirectly related (Online Resource 2), suggesting that Fe leaf nutrition under salinity (highly related to leaf [Ca2+]) could be a limiting factor for salt tolerance in citrus. The relationship of leaf water potential with traits in experiment 2 showed a major change as deduced from the comparison between the control and salinity graphic representations of principal component analysis (Fig. 4) and correlation analysis (Online Resource 2). Thus, while Ψ7 was mainly related to L-R_Na under control (Online Resource 2), the most significant traits associated with Ψ7 under salinity were R_B, R_Mo and R_Mn (Fig. 4 and Online Resource 2). The changes observed in all these physiological traits could be contributing to the salt tolerance of the progeny inherited from Cleopatra mandarin. How much genetic variation of these traits is available for breeding purposes and where is it located in the genome?

Genetic variation and QTL analysis of traits

A profitable amount of genetic variability has been found for several traits in the RxPr population (Fig. 2 and Online Resource 4). According to Tables 2 and 3, most trait heritabilities were below 0.6 although the estimates for some traits related to salt tolerance were ≥0.50 (Cl−, %DMA, DLW2, TDRW). This is not the case of leaf water potential, a usual indicator of xylem functionality in many drought and salinity tolerance studies (Pedroso et a. 2014; Rodriguez-Gamir et al. 2011; Forner-Giner et al. 2011; García-Sanchez et al. 2009; Ortuño et al. 2004; Morinaga and Sykes 2001; Savé et al. 1995) but whose heritability had not been previously estimated. Present results showed that its utilization for selection may show low predictive value, particularly, its late measure Ψ80 (Table 3).

The QTL analysis of the 60 traits evaluated under both environmental conditions, in both experiments, rendered a low number of significant QTLs. In fact, no QTL per trait was the mode class, particularly under control (Online Resource 4). Beyond the significance level fixed for the detection, the reduced number of QTLs detected per trait could be better explained by the reduced number of genotypes rather than by the trait heritabilities. Nevertheless, two facts of the present QTL analysis support a reasonable degree of reliability: (1) some QTLs of traits studied in both experiments showed repeatability, and (2) QTLs governing some traits under both control and salinity mapped together. Thus, from both linkage groups 3 and 7 involved in vegetative growth or plant vigour under control condition, those QTLs in linkage group 3 (Fig. 5) for FLW, DLW, LWC, TDRW and RWC were also detected under salinity and, importantly, in the same genotypic direction than under control, pointing out their relevance for selection in salt tolerance breeding programs of citrus rootstocks. Unfortunately, we cannot relate this region to that where Sahin-Çevik and Moore (2012) located QTLs for trunk diameter, tree high and tree canopy width in linkage group VII because of the lack of common markers.

Forty-five out of 98 QTLs (45.9 %) detected by using the IM procedure were coincident (markers in italics in Table 4) with the most significant markers associated by Kruskal-Wallis (KW, in bold in Online Resource 3). A similar proportion (52.9 %) was reported recently in another citrus progeny (Asins et al. 2015). The complementary nature of QTL information provided by both methods is also displayed in Fig. 5 for the most relevant linkage groups by specifying KW-associated traits in the correspondent integrated C. reshni-P. trifoliata linkage groups (7, 10+5b, 3, 4c). This representation is also helpful to connect trait correlations and linkage of QTLs from both parents governing the traits involved. The significant relationships among leaf traits (FLW, LWC, DLW, L_B) and among root traits (RWC, TDRW, R_Na) observed under salinity agreed with genotypic variation at associated markers in linkage group 3 (CL2.26,395, CMS30,160, CR80,260, C2iC11,470 and C8iC1rt,650 in Fig. 5, Table 4 and Online Resource 3). Noteworthily, three QTLs for the duration of the juvenility period (Jp3.2, Jp3.1 and Jp3.3) were previously reported in this same linkage group at C8intC1rt650, C2iC1i,470/TAA27 and CL2-26, respectively, in this segregant population (Raga et al. 2012) connecting genetically this trait to plant vigour, a well-known trait association in fruit trees (Visser 1970).

Salt tolerance and candidate genes

As far as salt tolerance is concerned, most QTLs for vegetative growth were detected in linkage group 3 (FLW, LWC, DLW, RWC, TDRW) where a L_B QTL also mapped. Given that they mapped together, is boron nutrition playing an active role in salt tolerance (citrus vigour under salinity)? We suspect it is not; i.e. it is a consequence of the gene substitution effects at these plant vigour QTLs and the salinity limiting effect on B nutrition. Thus, both L_B and R_B distributions moved to the left under salinity (Fig. 2h and l), but B is needed for root and stem growth (R_B_S and SD_S were correlated, Online Resource 2). In fact, the predominant function of B is in the formation of primary cell walls (O’Neill et al. 2004; Miwa and Fujiwara 2010; Wang et al. 2015). Therefore, less B would reach the leaves in large-rooted vigorous genotypes (genotypes with higher RWC_S), and this lower L_B_S would also explain a higher DI_S. These relationships agreed with the Cleopatra-allele substitution effects at CMS30,160 in linkage group 3 where QTLs for L_B, FLW2, DLW2, RWC and DI (Online Resource 3, Fig. 5) were detected under salinity. These results on the salt tolerance of rootstocks derived from Cleopatra mandarin might be useful in order to increase citriculture sustainability in the Mediterranean area because they provide a new perspective to the suitability of saline wastewaters (frequently containing elevated B concentration) for irrigation of citrus orchards (Grattan et al. 2015).

Two QTLs for leaf area growth (LAG) and leaf water potential-related traits were detected in linkage groups 7 and 4c (Fig. 5). These LAG QTLs corresponded to control, but none was detected under salinity. However, in the case of linkage group 7, two QTLs for LA under both control and salinity were also detected by KW in the same genomic region (Online Resource 3). Although the most significant marker is not the same for both LA_C and LA_S, if we compare genotypic means for Ψ80_S, DI_S, L_Na_S, LA_S (significant here only at p ≤ 0.05), LA_C and LAG_C at the CR17,300 marker in linkage group 7 (Fig. 6), the genotype bd showed both the largest L_Na_S and LA_S means. Besides, the inverse relationship observed between L_Na_S and DI_S (Online Resource 2) was supported by the genotypic means at this marker. These results suggest the presence of a putative leaf-Na+ tolerance QTL here. Regarding the cluster of QTLs for dΨ_S, L_K_S, R_Na_S and Cl_S in linkage group 4c (Fig. 5, Table 4 and Online Resource 3), the same marker locus CR23,750 was linked to a QTL for LAG_C (in the Poncirus genome) and another for Cl_S (in the Cleopatra genome), suggesting that the genotypic difference in Cl− accumulation might silence any difference between Poncirus alelles at the LAG QTL, what could be reasonable to interpret as a leaf-Cl− sensitivity QTL. Tozlu et al. (1999b) found a genomic region (NaCl1c) involved in Na+ and Cl− accumulation under salinity where growth and dried mass-related QTLs mapped (Tozlu et al. 1999a), but they did not associate it with tissue salt sensitivity. In fact, their linkage group I where they placed NaCl1c corresponded to our linkage group 8+6 (see location of common marker Got1 in Cai et al. 1994 and Ruiz and Asins 2003). Differences could be explained by the salt-tolerant donor used in each case.

Seven markers related to salt tolerance candidate genes (42C, COR15, Acuapor, SOS1, SOS2, CCC and Ethrec), one marker for the microARN miR164 and two markers for transcription factors (Mybg2 and EREBP1) were found linked to QTLs detected by IM and/or KW for traits evaluated under both control and salinity (Table 4 and Online Resource 3). Regarding candidate genes associated with salt tolerance rootstock effects in the C. volkameriana × P. trifoliata grafted population reported by Raga et al. (2014), linked markers to CCC and CNHX1 were associated to salt-tolerance-related traits in two cases: 3F-4R,900 (linked to CCCAlu,650 in p10+5b) associated with LWC1_S (and R_Mn_S) and CMS30,160 (linked to CNHX, 1600 in linkage group 3) associated with several LWC traits. No Na+ QTL was found associated with salt tolerance candidates SOS1, SOS2 or NHX1 but a L_Na_C QTL was detected at Acuapor,750, a marker derived from an aquaporin gene, in r4b. Leaf Cl− accumulation was not found associated with marker CCCAlu1,650 (corresponding to a citrus-chloride cotransporter reported by Brumos et al. 2009) in p10+5b; however, the maximum LOD score of QTLs for the root concentrations of Mn (trait significantly related to Ψ7_S), P and Ti under salinity located at this maker by IM (Table 4, Fig. 5). Regarding leaf water potential, two QTLs were detected for leaf water potential change under salinity (dΨ_S in Table 4) that were somehow related to C. tristeza virus (CTV) infection. One of those dΨ_S QTLs was detected at miR164 marker (in p4c) which corresponds to a microRNA gene from a CTV-challenged phloem cDNA library (Song et al. 2009). Jones-Rhoades et al. (2006) reported that the miR164 expression was high in leaf, shoot and root of Poncirus and its target genes were NAC transcription factors whose overexpression induces, among other effects, reduced lateral rooting. In the case of salinity, Barciszewska-Pacak et al. (2015) have shown recently that it increased Arabidopsis leaf miR164c-3p expression more than four times. The other QTL detected for dΨ _S corresponded to a candidate gene for CTV resistance (resistance analogue Py65 reported by Yang et al. 2001). Of course, plants of experiment 2 were obtained from seeds and kept under a greenhouse, so they were virus free. This possible connection between both biotic and abiotic stresses and the vascular system deserves a future genomic insight taking advantage of the genome sequences already available for Citrus at the web portals http://phytozome.jgi.doe.gov/and https://www.citrusgenomedb.org/.

In conclusion, salt tolerance in terms of vegetative growth, leaf water potential and nutrient distribution has been genetically studied using nucellar seedlings of a population derived from two common but far related citrus rootstocks, trifoliate orange (P. trifoliata) and Cleopatra mandarin (C. reshni), a very well-known salt tolerance rootstock whose inheritance had never been studied before. Correlation and QTL analyses of 60 traits in two experiments under both control and salinity conditions have shown that salt tolerance was mostly related to the whole plant vigour and its ionomic profile. The position of QTLs reported for plant vigour traits in linkage groups 3 and 7 might be useful to obtain selection tools to improve rootstocks for salt tolerance and as genomic reference points where to continue searching for candidate genes.