Inheritance of seed desiccation sensitivity in a coffee interspecific cross: evidence for polygenic determinism

Abstract

The genetic determinism of seed desiccation sensitivity was studied using a cross between two coffee species exhibiting a large difference for this trait, Coffea pseudozanguebariae (tolerant) and C. liberica (sensitive). Throughout the whole study, seed desiccation tolerance was quantified both in terms of water content and water activity. Whatever the parameter used, the level of seed desiccation tolerance in F₁ hybrids corresponded to that of the mid-parent, thus indicating an additive inheritance of seed desiccation tolerance at the F₁ level. A broad variation was observed among hybrids backcrossed to C. liberica (BCs) for seed desiccation tolerance, independent of the parameter used to quantify it. This variation was continuous and BCs showed transgression in the direction of the most desiccation sensitive parent, indicating (i) that desiccation tolerance is a polygenic trait in coffee species, and (ii) that C. pseudozanguebariae does not present the most favourable alleles for all the genes involved in seed desiccation tolerance. No significant difference was observed between the two reciprocal backcrosses, F₁×C. liberica and C. liberica×F₁, for the level of desiccation tolerance of their seeds, showing the absence of a maternal effect on this trait. There was no significant effect of the number of seeds harvested from each BC on the level of desiccation tolerance of its seeds. Moreover, there was no significant correlation within BCs between seed size, seed viability, and water content before desiccation and desiccation tolerance.

Additivity, Coffea, coffee, desiccation sensitivity, desiccation tolerance, intermediate seed, transgression

Introduction

Three categories of seed storage behaviour are generally recognized among plant species: orthodox, intermediate, and recalcitrant (Roberts, 1973; Ellis et al., 1990). This terminology is also frequently used to classify species depending on the level of desiccation tolerance of their seeds (Black and Pritchard, 2002): orthodox seeds tolerate the immediate effects of extreme water loss; intermediate seeds withstand considerable (down to water potentials of about −150 MPa) but not complete drying; recalcitrant seeds do not survive if desiccated to water potentials lower than −15 MPa. The subdivision of non-orthodox seed species in two categories does not reflect the very high variation that has been detected for seed desiccation sensitivity within both the intermediate and recalcitrant categories. For example, intermediate species of the genus Coffea display a broad variability in seed desiccation sensitivity (Dussert et al., 1999; Eira et al., 1999a). In a previous study that included nine coffee species, it was observed that the seed water content at which half of the initial viability was lost ranged from 0.05 g H₂O g⁻¹ DW with Coffea pseudozanguebariae to 0.38 g H₂O g⁻¹ DW with C. humilis (Dussert et al., 1999). This interspecific variability was broad enough to include species whose seeds were sufficiently desiccation-tolerant to be cryopreserved and others that could not (Dussert et al., 2001). Therefore, for some intermediate species, seed desiccation sensitivity may constitute a considerable problem for long-term germplasm conservation by means of seed cryopreservation.

Results obtained recently using neem (Sacandé et al., 2001) and coffee (Eira et al., 1999b) as model species revealed some of the constitutive characteristics of intermediate seeds (specific embryo water sorption properties; particular seed membrane properties). However, reasons for the high variability observed among intermediate species for the level of desiccation sensitivity of their seeds remain elusive. One approach to study this variability is to search for correlations between the level of desiccation tolerance and some biophysical or biochemical characteristics of the seeds. Di- and oligosaccharides have been proposed to play an important role in the range of hydration levels where coffee seeds are damaged by desiccation (Crowe et al., 1992; Hoekstra et al., 1997). However, the differences in seed desiccation tolerance observed between coffee species were not significantly correlated with differences in seed soluble sugar content, suggesting that sugars alone do not confer seed desiccation tolerance (Chabrillange et al., 2000).

Alternatively, some authors have proposed that the apparent variation for desiccation tolerance among non-orthodox seed species is only an artefact due to the use of water content as the variable to express seed water status (Sun and Liang, 2001; Walters et al., 2002). According to these authors, desiccation sensitivity of non-orthodox seed species falls into discrete levels when it is quantified in terms of water potential and not in terms of water content. For example, Sun and Liang (2001) proposed the existence of discrete levels at −23 and −73 and MPa for intermediate seed species. According to this hypothesis, desiccation sensitivity would not be a quantitative trait in itself, but would represent a series of challenges designed to meet at discrete water potentials (Walters et al., 2002).

Studying the genetic determinism of seed desiccation sensitivity using crosses between species/genotypes exhibiting a high difference for this trait represents an interesting approach to investigate the cause of (i) the interspecific variability observed for seed desiccation tolerance, and (ii) the existence of a continuum versus discrete levels of seed desiccation sensitivity. Such an approach has already been used, notably for studying variability in seed longevity in rice (Miura et al., 2002) and Arabidopsis thaliana (Bentsink et al., 2000) and for the genetic analysis of several compounds that contribute to coffee brew quality, such as seed chlorogenic acid (Campa et al., 2003) and trigonelline (Ky et al., 2000a) contents. In the present study, the inheritance of seed desiccation sensitivity was investigated using a cross between C. pseudozanguebariae and C. liberica var. dewevrei, two coffee species which display significant differences in seed desiccation sensitivity (Dussert et al., 1999; Eira et al. 1999a). This study provides some new results on the quantitative nature of desiccation sensitivity in coffee species and on the possible number of genes involved in this trait. Throughout the whole study seed desiccation tolerance was quantified both in terms of water content and water activity (function of water potential). Moreover, this study established the technological possibility of developing a candidate gene approach for a better understanding of desiccation sensitivity in intermediate seeds.

Materials and methods

Plant material

Plant material was maintained by and obtained from the IRD Agricultural Station at Man, Côte-d'Ivoire. In 2001, the inheritance of seed desiccation sensitivity was studied through the evaluation of seven genotypes of each of the two parental species, Coffea liberica var. dewevrei Hiern and C. pseudozanguebariae Bridson, seven F₁ hybrids, and 23 hybrids backcrossed to C. liberica (Table 1). Each genotype studied corresponded to one tree. The genotypes of C. liberica and C. pseudozanguebariae studied were wild accessions originating from the Central African Republic and Kenya, respectively. Interspecific F₁ hybrids were obtained from crosses using C. pseudozanguebariae as the female parent and C. liberica as the pollinator. The reciprocal cross never gave any progeny. Backcross hybrids (BCs), C. liberica×F₁ or F₁×C. liberica, originated from hand-pollinated crosses. The 23 BCs studied in 2001 were chosen among genotypes showing high fertility (producing more than 500 seeds). By contrast, the 58 BCs studied in 2002 were chosen randomly in the plot, independent of their fertility. The total number of seeds produced by each of these 58 genotypes was recorded as a fertility estimator.

Table 1.

List of the saturated salt solutions used for desiccation, and their corresponding relative humidity at 25 °C, and number of genotypes studied, n, for each type of plant material evaluated and each year of evaluation

Year	Plant material	n	Salts [relative humidity (%)]
2001	C. pseudozanguebariae	7	ZnCl2 [6], KOH [9], K acetate [23], MgCl2 [32], K2CO3 [45], NH4NO3 [62], NH4Cl [78]
	F1 hybrids	7
	Backcross hybrids	23	KOH [9], K acetate [23], MgCl2 [32], K2CO3 [45], NH4NO3 [62], NH4Cl [78], BaCl2 [90]
	C. liberica	7
2002	Backcross hybrids	58	K2CO3 [45]

Desiccation and germination procedures

In all experiments, the endocarp was manually removed from the seeds before treatment. For each genotype studied, seed viability and water content were measured immediately upon receipt in the laboratory. Seed viability at receipt, V_fresh, was estimated using two and three replications of 30 seeds in 2001 and 2002, respectively. For the measurement of seed desiccation sensitivity, in 2001, seeds were desiccated by equilibration at 25 °C over various saturated salt solutions as previously described by Dussert et al. (1999). The list of salts used for parents, F₁ hybrids, and BCs is given in Table 1. In 2002, seeds were desiccated by equilibration over a saturated K₂CO₃ solution only. Seed viability after desiccation was estimated using two and three replications of 30 seeds per desiccation condition in 2001 and 2002, respectively. The water content (WC) of seeds at equilibrium (expressed in g H₂O g⁻¹ DW) was always estimated using 10 replicates of one seed and their dry weight was measured after 1 d of desiccation in an oven at 105 °C. For germination, batches of 10 seeds were placed over 18 g vermiculite fully imbibed with 45 ml sterile water in closed plastic boxes (©Magenta). Seed viability was assessed using the criterion of normal seedling development, i.e. emergence of the hypocotyl, radicle geotropic growth (>20 mm), and the opening of cotyledonary leaves, after 6 weeks of culture at 25 °C in the dark.

Quantification of desiccation sensitivity

For the 44 genotypes studied in 2001, quantification of desiccation sensitivity was performed using the quantal response models developed previously (Dussert et al., 1999, 2003) which estimate the water content, WC₅₀, and the water activity, a_w50, at which half of the initial viability is lost. The water potential at which half of the initial viability is lost, Ψ₅₀, was calculated from a_w50 using the equation Ψ=RTln(a_w)/V_w (Pa), with R=8.314, T=298 K, and V_w=18.07×10⁻³ l. For the 23 back-crossed genotypes studied in 2001, seed desiccation sensitivity was also quantified by a new parameter, Survival_45%RH, which corresponded to the percentage of seeds developing into normal seedlings after equilibration in a 45% relative humidity atmosphere divided by V_fresh. Survival_45%RH was also used to quantify seed desiccation sensitivity in the 58 BCs studied in 2002.

Results

Inheritance of the level of seed desiccation tolerance in F₁ hybrids

For each of the 44 genotypes studied in 2001, whatever the variable used for describing seed hydration level (water content or water activity), the relationship between seed hydration level and their viability followed the typical sigmoidal pattern expected (Fig. 1). In all cases, the proportion of variance explained by the model was very high (mean, minimum, and maximum values of R² were 0.98, 0.92, and 0.99, respectively).

Fig. 1.

Quantification of seed desiccation sensitivity using the quantal response model: examples of application to one genotype of C. pseudozanguebariae (open circles), one genotype of C. liberica (open triangles) and one F₁ hybrid (open squares) evaluated in 2001.

Seeds of the seven C. pseudozanguebariae genotypes exhibited a very high level of tolerance to desiccation, since the mean values of WC₅₀ and a_w50 were 0.055 g H₂O g⁻¹ DW and 0.135, respectively (Fig. 2). The level of seed desiccation sensitivity observed in the seven genotypes of C. liberica was significantly higher (P<0.0002) with mean values of WC₅₀ and a_w50 of 0.113 g H₂O g⁻¹ DW and 0.413, respectively. In both parental species, the standard deviation observed for the level of seed desiccation tolerance was very low, indicating a very low intraspecific diversity for this trait.

Fig. 2.

Levels of seed desiccation sensitivity observed in parental species and F₁ hybrids and regression analysis showing additive inheritance. Seed desiccation sensitivity was quantified by the water content and the water activity at which half of the initial viability was reached, WC₅₀ (A) and a_w50 (B), respectively. Ψ₅₀ is the calculated water potential corresponding to a_w50.

Whatever the parameter used, WC₅₀ or a_w50, the level of seed desiccation tolerance in F₁ hybrids was significantly different (P<0.0003) from those of the parental species and corresponded with that of the mid-parent, i.e. the mean of the values observed in the two parental species (Fig. 2). The linear regression performed using values of the 21 genotypes of C. pseudozanguebariae, C. liberica, and the F₁ hybrids studied was highly significant (r=0.95; P=0.0000) for both variables. These results showed an additive inheritance of seed desiccation tolerance at the F₁ level.

Variation of the level of seed desiccation tolerance in second-generation hybrids

A large variation was observed within the 23 BCs studied in 2001 for the level of desiccation tolerance of their seeds, independent of the parameter used to quantify it (Fig. 3). WC₅₀ ranged between 0.086 and 0.145 g H₂O g⁻¹ DW, thus from values observed in F₁ hybrids up to values considerably higher than those found in C. liberica. The same phenomenon was observed for a_w50 with values ranging between 0.25 and 0.60. Therefore, BCs showed transgression in the direction of the most desiccation-sensitive parent.

Fig. 3.

Distribution of the level of seed desiccation sensitivity in the 23 backcross hybrids evaluated in 2001. Seed desiccation sensitivity was quantified by the water content and the water activity at which half of the initial viability was reached, WC₅₀ (A) and a_w50 (B), respectively. Arrows indicate the mean values observed in the parental species and the F₁ hybrids for the two parameters used for desiccation sensitivity quantification. Ψ₅₀ is the calculated water potential corresponding to a_w50.

Because of the wide range of values observed in BCs, the number (23) of genotypes studied in 2001 appeared to be too low to determine whether desiccation tolerance followed a monomodal or a plurimodal distribution. A larger population of BCs was thus employed in 2002. A simplified method was developed to measure seed desiccation tolerance in such a large population. All correlations between either WC₅₀ or a_w50 and the viability percentages (expressed in percentage of the control, V_fresh) obtained with the different RHs used were tested. The best correlation was obtained with 45% RH, a result that was expected since the mean value for a_w50 in the 23 BCs studied in 2001 was about 0.45. The correlation between Survival_45%RH and either WC₅₀ or a_w50 was highly significant and Survival_45%RH thus constituted a very accurate simplified estimator of the level of seed desiccation tolerance in BCs (Fig. 4).

Fig. 4.

Relationship between the proportion of seeds developing into normal seedlings after equilibration under 45% RH and the water content and the water activity at which half of the initial viability was reached, WC₅₀ (A) and a_w50 (B), respectively, within the 23 backcross hybrids evaluated in 2001.

Survival_45%RH was measured in 2002 in 58 BCs. Again a very large variation for seed desiccation tolerance was observed in BCs (Fig. 5). Survival_45%RH ranged from 7.6–100%. Transgression in the direction of the most sensitive parent was confirmed, since more than half of the BCs showed a desiccation tolerance lower than C. liberica genotypes. The variation for desiccation tolerance was continuous in BCs, i.e. no discrete levels could be identified.

Fig. 5.

Distribution of the level of seed desiccation sensitivity in the 58 backcross hybrids evaluated in 2002. Seed desiccation sensitivity was quantified by the proportion of seeds developing into normal seedlings after equilibration under 45% RH. Arrows indicate the mean values observed in the parental species and the F₁ hybrids for the parameter used for desiccation sensitivity quantification.

Correlation between levels of desiccation tolerance observed in 2001 and 2002

Among all BCs studied in 2002, 14 were also evaluated for the level of desiccation tolerance of their seeds in 2001. Within these 14 genotypes, a significant correlation was observed between Survival_45%RH data obtained in 2002 and WC₅₀ (r=−0.69, P=0.0065) and a_w50 (r=−0.62, P=0.0241) values estimated in 2001.

Testing the hypothesis of a maternal cytoplasmic effect on seed desiccation tolerance

Whatever the variable used for quantifying desiccation tolerance, WC₅₀ and a_w50 for the genotypes studied in 2001 and Survival_45%RH for those analysed in 2002, no significant difference was observed between the two reciprocal backcrosses, F₁×C. liberica and C. liberica×F₁, for the level of desiccation tolerance of their seeds (Table 2). This result shows the absence of a maternal cytoplasmic effect on the level of seed desiccation tolerance.

Table 2.

Comparison of the mean level of seed desiccation sensitivity within the reciprocal backcrosses F₁×C. liberica and C. liberica×F₁

	Year 2001					Year 2002
	n	WC50 (g g−1)	aw50	n	Survival45%RH (%)
F1×C. liberica	15	0.1098	0.3974	46	51.74
C. liberica×F1	8	0.1178	0.4438	12	62.15
F		1.3533	1.6299		1.1798
P		0.2577	0.2156		0.2818

Seed desiccation sensitivity was quantified by the water content and the water activity at which half of the initial viability was reached, WC₅₀ and a_w50, respectively, in 2001, and by the percentage of seeds developing into normal seedlings after equilibration under 45% RH in 2002. Results of one-way ANOVAs: F and P. Number of genotypes studied per cross and per year: n.

Influence of initial seed viability and water content, seed size, and tree fertility on the level of desiccation tolerance

There was no significant (P>0.05) effect of the number of seeds harvested from each genotype studied in 2002 on the level of desiccation tolerance of its seeds (Table 3). This result precludes any bias in the measurement of desiccation tolerance caused by differences in fertility of BCs. Moreover, there was no significant (P>0.05) correlation between seed viability and water content before desiccation (upon receipt in the laboratory) and desiccation tolerance. Thus, post-harvest environmental factors did not greatly influence the trait studied. Seed size, as estimated by seed dry weight, was not significantly (P>0.05) correlated to desiccation tolerance. This result suggests that the initial desiccation rate (during the first week of equilibration) did not influence significantly the estimation of the level of desiccation tolerance.

Table 3.

Mean, standard deviation (SD), minimum (Min), and maximum (Max) values of seed viability and water content (WC) at receipt in the laboratory, total number of seeds harvested and seed dry weight in the 58 backcross hybrids evaluated in 2002 and probability (P) of correlation with seed desiccation sensitivity, as quantified by the percentage of seeds developing into normal seedlings after equilibration under 45% RH

	Mean	SD	Min	Max	P
Seed viability at receipt (%)	84.1	11.6	66.6	97.9	0.8029
Seed WC at receipt (g g−1)	0.672	0.111	0.473	0.922	0.9468
Number of seeds harvesteda	431.7	217.1	120	1000	0.5598
Seed dry weight (mg)	88.7	24.3	44.8	155.7	0.7523

a

For trees which have produced more than 1000 seeds, the value of 1000 was recorded for the total number of seeds harvested.

Discussion

The present study provides valuable and original information on the nature of seed desiccation tolerance in coffee species by using a cross between two coffee species showing a large difference in the level of seed desiccation tolerance and a very low intraspecific variability for this trait. To the authors' knowledge, it is the first time that a genetic approach has been used to study seed desiccation tolerance in non-orthodox seed species. This study clearly shows the quantitative nature of desiccation tolerance in seeds belonging to this category of storage behaviour. When the level of desiccation tolerance is expressed in terms of water potential, the additivity observed in F₁ hybrids (−178 MPa, equivalent to mid-parent) and the large segregation in BCs (−189 to −68 MPa) are not consistent with the hypothesis of the existence of discrete levels of seed desiccation tolerance within the intermediate category (Sun and Liang, 2001). Interestingly, seeds of the cultivated species C. arabica, which is an allotetraploid, show a level of desiccation tolerance intermediate between those of its two parental (genome donor) species, C. eugenioides and C. canephora (Dussert et al., 1999). This observation, which is consistent with the additivity observed in C. pseudozanguebariae×C. liberica F₁ hybrids, supports the hypothesis that speciation leads to a continuum of desiccation tolerance levels among non-orthodox species (Pammenter and Berjak, 1999). Moreover, the existence of such a continuum does not contradict the proposal that desiccation induces the same type of damage in intermediate seeds of different plant species (e.g. Citrus sp., Coffea sp., neem, oil palm, …) (Walters et al., 2002). Sacandé et al. (2001) proposed that the intermediate storage behaviour of neem seeds and, in particular, their sensitivity to rapid rehydration when desiccated, is due to the intrinsically high gel–liquid crystalline temperature (T_m) of their membranes. It has recently been observed that all seed rehydration procedures (osmoconditioning, preheating, prehumidification) previously identified as reducing membrane injury (Woodstock and Tao, 1981; Hoekstra and van der Wal, 1988; Sacandé et al., 2001) dramatically increased the level of desiccation tolerance of C. arabica seeds, since a_w50 decreased from 0.41 down to 0.16 (Dussert et al., 2003). Assuming that desiccation-induced phase transition is the common starting point of a series of phenomena that leads to membrane dysfunctioning upon rehydration in intermediate species, it can be hypothesized that all factors identified so far which influence membrane T_m, including phospholipid acyl length and degree of unsaturation, phospholipid head group composition, sucrose content, amount of certain amphiphilic substances (Quinn, 1985; Hoekstra et al., 1992; Crowe et al., 1992; Hoekstra and Golovina, 2000), can contribute to the very large variation observed between intermediate species for seed desiccation tolerance. Differences in pollen WC₅₀ values (from 0.05 to 0.14 g H₂O g⁻¹ DW at 15 °C) observed by Hoekstra and van der Wal (1988) between different plant species, were shown to be significantly correlated to differences in membrane T_m, which were themselves correlated with pollen sucrose content and degree of unsaturation of phospoholipids (Hoekstra et al., 1992).

The occurrence of transgressive segregation in second-generation hybrids, independent of the variable used for quantifying desiccation tolerance, WC₅₀, a_w50 or Survival_45%RH, provides useful information on its genetic determinism in coffee species. It shows firstly that it is a polygenic trait, a fact which is in agreement with the common understanding that seed desiccation tolerance is a multi-factorial trait (see Buitink et al., 2002, for a review). Secondly, it shows that C. pseudozanguebariae does not present the most favourable alleles (i.e. that increase tolerance) for all the genes involved in seed desiccation tolerance, since some BCs displayed combinations of alleles less favourable than the most desiccation-sensitive parent, C. liberica. This result may explain why no significant correlation was found between seed sucrose content alone and desiccation tolerance in coffee species (Chabrillange et al., 2000). Seed sucrose content is higher in C. pseudozanguebariae than in C. liberica (Ky et al., 2000b; Chabrillange et al., 2000). Using the crosses employed in the present study, Ky et al. (2000b) showed that seed sucrose content exhibited an additive inheritance both in F₁ hybrids and in genotypes backcrossed to C. liberica. Therefore, it can be speculated that genes for which C. pseudozanguebariae present unfavourable alleles are not those controlling seed sucrose content.

In addition to demonstrating that the studied BCs segregated for desiccation tolerance, this study also rules out different potential drawbacks for support of a candidate gene approach of desiccation tolerance. Firstly, it established the absence of any maternal cytoplasmic effect on this trait. Secondly, the influence of BC fertility and initial seed viability and water content on desiccation tolerance was shown to be non-significant.

From a technological standpoint, it was difficult to evaluate a large number of BCs during the same year using the method previously developed for quantifying seed desiccation tolerance (Dussert et al., 1999) for the following reasons: (i) evaluation cannot be delayed after harvest; (ii) seed maturity among BCs is reached within a very short period of time; and (iii) some BCs showed low fertility and produced fewer seeds than requested. Therefore, the simplified method developed in the present study for estimating seed desiccation tolerance in BCs represents a significant methodological advance, since an accurate estimate was achieved using a reduced number of seeds. This method should be easily adaptable to any population of second-generation hybrids segregating for this trait by choosing an equilibration RH close to the median a_w50 value of the studied population. This simplified method is currently being used for mapping QTLs controlling seed desiccation tolerance in a large population of C. pseudozanguebariae×C. liberica BCs.