Exogenous gibberellins inhibit coffee (Coffea arabica cv. Rubi) seed germination and cause cell death in the embryo

Abstract

The mechanism of inhibition of coffee (Coffea arabica cv. Rubi) seed germination by exogenous gibberellins (GAs) and the requirement of germination for endogenous GA were studied. Exogenous GA₄₊₇ inhibited coffee seed germination. The response to GA₄₊₇ showed two sensitivity thresholds: a lower one between 0 and 1 μM and a higher one between 10 and 100 μM. However, radicle protrusion in coffee seed depended on the de novo synthesis of GAs. Endogenous GAs were required for embryo cell elongation and endosperm cap weakening. Incubation of coffee seed in exogenous GA₄₊₇ led to loss of embryo viability and dead cells were observed by low temperature scanning microscopy only when the endosperm was surrounding the embryo. The results described here indicate that the inhibition of germination by exogenous GAs is caused by factors that are released from the endosperm during or after its weakening, causing cell death in the embryo and leading to inhibition of radicle protrusion.

β-mannosidase, cell death, coffee seed, endo-β-mannanase, germination, gibberellins, puncture force

Introduction

Gibberellins (GAs) play an important role in the stimulation of seed germination (Bewley, 1997). GA-deficient mutants of Arabidopsis and tomato do not germinate in the absence of exogenous GA (Koornneef and van der Veen, 1980; Groot and Karssen, 1987). Tetcyclacis and paclobutrazol are inhibitors of GA-biosynthesis and may prevent seed germination (Karssen et al., 1989; Rademacher, 2000). Addition of exogenous GAs completely reverts the inhibitory effect of tetcyclacis and paclobutrazol, for example, in Arabidopsis, indicating that side-effects are absent (Debeaujon and Koornneef, 2000). GAs can promote germination of dormant seeds by their ability to overcome or ‘short-circuit’ the requirement for environmental factors that are required for germination, including afterripening, light and cold. This has led to the hypothesis that such environmental factors may induce GA biosynthesis during the early phases of germination (Hilhorst and Karssen, 1992). Indeed, it was shown that red light enhances GA₁ levels in photoblastic lettuce seeds (Toyomasu et al., 1993).

GAs may induce endosperm degradation by stimulating hydrolytic activity in the endosperm cell walls. This was first demonstrated in celery (Jacobsen et al., 1976) and pepper seeds (Watkins and Cantliffe, 1983; Watkins et al., 1985). In tomato seeds, GAs liberated from the embryo triggered weakening of the endosperm cap opposing the radicle tip, induced degradation of the endosperm cell walls, and allowed radicle protrusion (Groot and Karssen, 1987; Groot et al., 1988). Activities of endo-β-mannanase, β-mannosidase, and β-galactosidase, all involved in the hydrolysis of galacto-mannans, were enhanced in the endosperm of the GA-deficient (gib1) tomato mutant treated with exogenous GA₄₊₇. In the absence of GAs only α-galactosidase could be detected but no endo-β-mannanase and β-mannosidase (Groot et al., 1988). Also in Datura ferox endo-β-mannanase and β-mannosidase were induced by GA in the micropylar endosperm (Sánchez and de Miguel, 1997). In tobacco seeds GA₄ induced β-1-3 glucanase activity in the micropylar endosperm, which corresponded with endosperm rupture (Leubner-Metzger et al., 1996).

Besides promoting endosperm weakening, GA stimulated embryo growth in tomato possibly by enhancing the embryo growth potential (Karssen and Laçka, 1986; Karssen et al., 1989). GAs stimulate elongation in hypocotyls of dark-grown lettuce seedlings (Katsu and Kamisaka, 1981) and in Arabidopsis GA controls cell elongation in light- and dark-grown hypocotyls (Cowling and Harberd, 1999). Evidence is accumulating that expansins (EXP) are regulated by GA and ABA and, hence, are potential candidates for hormone-regulated cell expansion and embryo growth potential during germination. For example, in tomato seeds, LeEXP8 and LeEXP10 are specifically expressed in the embryo (Chen et al., 2001). However, an increase in embryo pressure potential (turgor) during GA-controlled germination has yet to be demonstrated.

Contrary to many reports on the stimulatory effect of GA during seed germination and tissue elongation, GA₃ inhibited radicle protrusion (Valio, 1976; Takaki and Dietrich, 1979; Takaki et al., 1979) and seedling emergence in coffee seed (Maestri and Vieira, 1961). This inhibition was proposed to be caused by mannose, a degradation product of the hydrolysis of mannans (Takaki and Dietrich, 1980). Coffee endosperm cell walls are composed mainly of mannans (Wolfrom et al., 1961). Mannose has been shown to inhibit ATP synthesis and normal hexose metabolism (Herold and Lewis, 1977) and it caused DNA laddering in Arabidopsis roots and maize suspension cells (Stein and Hansen, 1999) and also plays a role in gene regulation during photosynthesis (Jang and Sheen, 1994).

In the present study an attempt is made to unravel further both the mechanism of promotion and inhibition of coffee seed germination by GAs.

Materials and methods

Seed source

Coffee seeds from Coffea arabica L. cultivar Rubi were harvested in Lavras, MG, Brazil in 2000. The fruits were mechanically depulped, fermented, and the seeds were dried to 12% moisture content (fresh weight basis) and shipped to The Netherlands where they were stored at 10 °C.

Germination conditions

Seed coats were removed by hand and seeds were surface-sterilized in a 1% sodium hypochlorite solution for 2 min. Seeds were rinsed in water and imbibed in 10 ml of demineralized water or GA₄₊₇ (Sigma, St Louis, Mo, USA; minimum 90% GA₄ and <5% GA₇₊) solution in a concentration of 1000 μM, 100 μM, 10 μM, or 1 μM GA₄₊₇. GA₃ (Sigma) was also tested, but was approximately 10 times less effective than GA₄₊₇ (data not shown). The concentrations of paclobutrazol and tetcyclacis used were 400 μM, 300 μM, 200 μM, and 100 μM, 50 μM or 10 μM, respectively. Seeds were placed in 8.5 cm Petri dishes on two layers of filter paper no. 3 (Whatman, Maidstone, UK). During imbibition seeds were kept at 30±1 °C in the dark (Huxley, 1965; Valio, 1976; da Silva, 2002) or in continuous light (irradiance 4.8×10 μmol m⁻² s⁻¹). A stock solution of 10⁻³ M GA₄₊₇ was made in 1 N KOH, followed by adjusting the pH to 7.0 with 1 N HCl. As a control coffee seeds were germinated in a 20 mM KCl solution to eliminate possible osmotic side-effects by KOH or HCl. ATP (0.1, 1.0, and 10 mM) and 20 mM of P_i (KH₂PO₄+K₂HPO₄) were used as a source of organic and inorganic phosphate, respectively, added to the GA₄₊₇ solution. Stock solutions of 10⁻³ M tetcyclacis (a gift from BASF, Germany) and paclobutrazol (a gift from Syngenta, Enkhuizen, The Netherlands) were made in acetone (0.3% v/v) by vigorous stirring overnight. Preliminary experiments showed that the amount of acetone used did not have any effect on coffee seed germination (data not shown). The germination percentage was recorded daily until the number of seeds showing radicle protrusion was constant.

Water potential measurements

The water potential (ψ) and osmotic potential (ψ_π) of coffee embryos were measured by using a calibrated thermocouple psychrometer (Model HR-33T, Wescor, USA) with a C-52 sample chamber (Wescor, USA). Samples were equilibrated for 40 min and two readings were taken before starting the experiments to ensure that equilibrium had been attained. Cooling time was 45 s. The C-52 chamber was placed in an airtight glove box kept at 100% relative humidity by a stream of water-saturated air at a constant temperature of 25±1 °C. Embryos were isolated as described below and placed in the C-52 chamber. Three replications of 5 embryos each were used for the measurements. After measurement of the water potential the embryos were put in liquid nitrogen for the determination of osmotic potential. After 2 h in liquid nitrogen the embryos were left to thaw and the osmotic potential was determined. The pressure potential (ψ_p) was calculated from the equation: ψ_p=ψ–ψ_π.

Embryo growth

Embryos from 20 coffee seeds were isolated by cutting the endosperm with a razor blade. Embryo length was measured with calipers. After length measurement the embryos were separated into embryonic axes and cotyledons of which the lengths were measured as well. Alternatively, embryos were isolated and incubated on Murashige and Skoog medium (ICN Biomedicals, Ohio, USA) solidified with 7g l⁻¹ of agar. After autoclaving the medium was supplemented with 100 μM GA₄₊₇ or sugars (10 mM of mannose, glucose, galactose, fructose). The Petri dishes with embryos were incubated at 30 °C in the dark. After 10 d of incubation the total length of the embryos was measured.

Puncture force measurement

The required puncture force of individual endosperm caps was measured as described before (Toorop et al., 2000). Briefly, an S 100 material tester (Overload Dynamics Inc., Schiedam, The Netherlands) was used with a JP 50 load cell (Data Instruments Inc., Lexington, MA, USA) at a range of up to 10 lb. A probe with a hemispherical tip and a diameter of 0.7 mm was placed on the load cell. Endosperm caps were cut from the seeds and the embryo was removed. The endosperm cap was placed on the probe and perforated by moving the probe down into a polyvinyl chloride block with a conic hole with a minimum diameter of 1 mm. The force required to puncture the endosperm cap was used as a parameter for the mechanical restraint of this tissue. All data points are the average of 30 measurements.

Measurement of endo-β-mannanase activity

Extracts were made from 10 endosperm caps taken from seeds at different imbibition intervals. Endo-β-mannanase was extracted in McIlvaine buffer (0.05 M citrate/0.1 M Na₂HPO₄, pH 5.0) with 0.5 M NaCl and assayed in a gel (0.5 mm thick) containing 0.5% (w/v) locust bean gum (Sigma) in McIlvaine buffer (pH 5.0) and 0.8% type III-A agarose (Sigma) on Gelbond film (Pharmacia). Samples (2 μl) were applied to holes punched in the gel with a 2 mm paper punch. Gels were incubated for 21 h at 25 °C, washed in McIlvaine buffer (pH 5.0) for 30 min, stained with 0.5% (w/v) Congo Red (Sigma) for 30 min, washed with ethanol for 10 min, and destained in 1 M NaCl overnight. Commercial endo-β-mannanase from Aspergillus niger (Megazyme, Cork, Eire) was used to generate a standard curve. Calculation of enzyme activity in the samples was according to Downie et al. (1994).

β-mannosidase extraction and assay

Ten endosperm caps (0.1 g) were ground in a mortar with liquid nitrogen. Enzyme was extracted from the seed parts with McIlvaine buffer pH 5.0 with 0.5 M NaCl. The extracts were centrifuged for 20 min at 21 000 g at 4 °C. Enzyme activity in the supernatant was assayed using 75 μl MacIlvaine buffer pH 5.0, 15 μl 10 mM p-nitrophenyl-β-D-mannopyrannoside (Boehringer, Mannheim) dissolved in MacIlvaine buffer, pH 5.0 and 60 μl enzyme extract. After incubation for 2 h at 37 °C the reaction was stopped by adding 75 μl 0.2 M Na₂CO₃. The yellow colour produced was measured at OD₄₀₅ in a microtitre plate reader. The enzyme activity was expressed as p-nitrophenol released (nmol s⁻¹) per endosperm cap.

Tetrazolium stain

Embryos were isolated and incubated in 0.1% (w/v) 2,3,5-triphenyltetrazolium chloride (Sigma) at 30 °C in the dark for 16 h, according to Dias and Silva (1986). The tetrazolium salts were used to measure the activity of dehydrogenase enzymes as an index of the respiration rate and seed viability, distinguishing between viable and dead tissues (Copeland and McDonald, 1999).

Cryo-scanning electron microscopy

Seeds from water-, GA₄₊₇- and mannose-imbibed seeds were prepared for cryo-scanning electron microcopy (cryo-SEM). The embryos were mounted on aluminium rivets with a drop of tissue-freezing medium (Tissue Tek, Sakura, The Netherlands). After mounting, the samples were plunge-frozen in liquid propane and stored in liquid nitrogen for subsequent cryo-planing and observations. The embryos were cryo-planed at −90 °C in a cryo-ultramicrotome (Reichert-Jung Ultracut E/FC4D) with a diamond knife (8 mm wide; 45°, Drukker International, The Netherlands) according to Nijsse and van Aelst (1999). The planed surfaces were freeze-dried for 3 min at −89 °C and 10⁻⁴ Pa and sputter-coated with platinum in an Oxford CT1500 HF cryo transfer unit. The surfaces were photographed in a cryo-SEM (JEOL 6300 F) at −180 °C and 2.5−5.0 kV using a digital imaging system.

Statistical analysis

Statistical analyses were performed by using the general linear model (SPSS 10.0.5).

Results

Radicle protrusion of the first seed occurred at day five of imbibition in water and was partially inhibited by light (Fig. 1). GA₄₊₇ inhibited radicle protrusion in a concentration-dependent manner. However, the dose–response relationship was not linear or log-linear as expected. Rather, a two step inhibition was observed with high and low sensitivity thresholds: 1 and 10 μM GA₄₊₇ resulted in a reduction of the maximal germination by 35% whereas 100 and 1000 μM GA₄₊₇ led to a further reduction by 30%. Apparently, the concentration thresholds for inhibition were between 0 and 1 μM and 10 and 100 μM, respectively.

Fig. 1.

Germination of coffee seeds in water, in the dark or light, and in various concentrations of GA₄₊₇ in the dark. Data points are averages of four replications of 25 seeds each; error bars indicate standard deviation.

The inhibition of radicle protrusion by exogenous GAs was only observed in coffee seeds. The same GA₄₊₇ solution was used to germinate tomato seeds and no inhibition of germination was observed, but GA₄₊₇ increased the speed of radicle protrusion (results not shown). Tetcyclacis and paclobutrazol, inhibitors of GA biosynthesis (Rademacher, 2000), completely inhibited germination at concentrations of approximately 400 μM and 300 μM, respectively (Fig. 2). Application of exogenous GA₄₊₇ overcame the inhibition and allowed germination, which excluded the possibility of side-effects during imbibition, and confirmed the requirement for GA-biosynthesis of coffee seed germination (Fig. 3). The dose–response curve displayed a narrow optimum of approximately 2 μM of GA₄₊₇ at the paclobutrazol concentration used. Germination in ATP or KH₂PO₄+K₂HPO₄ did not overcome the inhibitory effect of exogenous GA₄₊₇ (Fig. 4).

Fig. 2.

Germination of coffee seeds in paclobutrazol and in tetcyclacis. Data points are averages of four replications of 25 seeds each; error bars indicate standard deviation.

Fig. 3.

Germination of coffee seeds in various GA₄₊₇ concentrations in the presence of 300 μM paclobutrazol. Data points are averages of four replications of 25 seeds each; error bars indicate standard deviation.

Fig. 4.

Germination of coffee seeds in 0.1, 1 and 10 mM ATP and 20 mM of P (KH₂PO₄+K₂HPO₄) in the presence of 100 μM of GA₄₊₇. Data points are averages of four replications of 25 seeds each; error bars indicate standard deviation.

Water relations

Psychrometric measurements were started after 2 d of imbibition. At this time water uptake was not yet completed; this occurred around day 3 (results not shown). At 2 d of imbibition in GA₄₊₇ the embryo water potential was −7.7 MPa and increased to −4.3 MPa at day 4 of imbibition; the osmotic potential increased from −8.2 MPa to −5.7 MPa. This resulted in a significant increase in pressure potential from 0.4 to 1.4 MPa. At day 5 of imbibition a decrease in embryo water and osmotic potential was observed. The water potential decreased from −4.3 to −5.8 MPa, the osmotic potential from −5.7 MPa to −6.3 MPa and the pressure potential decreased significantly from 1.4 to 0.4 MPa. From 5 d of imbibition onwards the water-, osmotic- and pressure potential increased again (Fig. 5A).

Fig. 5.

Water potential ψ (filled circles), osmotic potential ψ_π (open circles) and pressure potential Ψ_p (filled inverted triangles) of coffee embryos isolated from seeds imbibed in 100 μM of GA₄₊₇ (A), 400 μM of tetcyclacis (B), or 300 μM of paclobutrazol (C). Data points are the averages of three replications of five embryos each; error bars indicate standard deviation.

The embryo water potential in tetcyclacis- and in paclobutrazol-imbibed seeds increased from −8.1 and −8.3 MPa to −3.1 and −4.2 MPa at day 7 of imbibition, respectively. The osmotic potential increased from −8.3 at day 2 to −3.2 MPa at day 7 in tetcyclacis-imbibed seeds. For paclobutrazol-imbibed seeds the osmotic potential increased from −8.3 at day 2 to −4.8 MPa at day 7 of imbibition. There was no significant increase in pressure potential in tetcyclacis- and in paclobutrazol-imbibed seeds (Fig. 5B, C).

Embryo growth

The embryo grew inside the endosperm during germination in GA₄₊₇-, tetcyclacis- and paclobutrazol-imbibed seeds until 10 d of imbibition when 50% of the seed population had germinated in non-inhibiting conditions. In GA₄₊₇-imbibed seeds the embryo grew approximately 35%, which was mainly due to elongation of the axis. In water-imbibed seeds the embryo showed growth to the same extent (Fig. 6A; da Silva et al., 2002). In paclobutrazol- and in tetcyclacis-imbibed seeds the embryo extended about 20% up to day 4 of imbibition and growth levelled off thereafter (Fig. 6B, C). Embryos incubated in 10 mM of mannose, glucose, galactose, or fructose showed a similar final growth as the water control (Fig. 7).

Fig. 6.

Length of the embryo (filled circles), axis (open circles), and cotyledons (filled inverted triangles) from coffee seeds imbibed in 100 μM GA₄₊₇ (A), 300 μM paclobutrazol (B), or in 400 μM of tetcyclacis (C). The embryos were isolated shortly before measurements. Data points are averages of 20 embryos each; error bars indicate standard deviation.

Fig. 7.

Embryo length in water, 10 mM fructose (fru), galactose (gal), glucose (glu), and mannose (man) or 100 μM of GA_{4+7 (}GA). The embryos were isolated after 1 d of imbibition in water, placed on Murashige and Skoog medium solidified with 7 g l⁻¹ of agar. Measurements were taken after 10 d of incubation at 30 °C in the dark. Data points are averages of 20 embryos each; error bars indicate standard deviation.

Puncture force measurement

The required puncture force was measured in the endosperm cap of coffee seeds imbibed in water, GA₄₊₇, tetcyclacis or paclobutrazol. In water, there was a significant decrease in the required puncture force prior to radicle protrusion. It was 1.37 N at 2 d of imbibition and decreased to 0.50 N at 10 d of imbibition (Fig. 8). In tetcyclacis-imbibed seeds the required puncture force was 1.20 N at 2 d of imbibition and it decreased to 0.80 N at day 10 after which it remained constant. Paclobutrazol imbibed seeds showed similar results as the tetcyclacis-imbibed seeds. The puncture force declined from 1.32 N at day 2 of imbibition to 0.85 at day 10. In GA₄₊₇-imbibed seeds the puncture force showed a significant decrease in the required puncture force to 0.55 N at 8 d of imbibition (P <0.001), after which it levelled off (Fig. 8).

Fig. 8.

The required puncture force of water-, tetcyclacis-, GA₄₊₇-, and paclobutrazol-imbibed seeds before radicle protrusion. Data points are averages of 30 measurements; error bars indicate standard error of mean.

Endo-β-mannanase activity

Parallel with the puncture force measurements endo-β-mannanase activity was determined in the endosperm caps. In seeds imbibed in water endo-β-mannanase activity was detectable from 2 d of imbibition onwards. The activity increased before the first seed completed germination at day 5 (Fig. 9A). In GA₄₊₇-imbibed seeds endo-β-mannanase activity in the endosperm cap was substantially higher (3–10-fold) than in water-imbibed seeds until 8 d of imbibition. After that, the activity decreased to the level of water-imbibed seeds at day 10 and was maintained at that level in the non-germinating seeds. Endo-β-mannanase activity was almost completely inhibited in tetcyclacis and paclobutrazol-imbibed seeds at all imbibition intervals.

Fig. 9.

Endo-β-mannanase (A) and β-mannosidase (B) activities in water, 100 μM of GA₄₊₇, 300 μM of paclobutrazol, and 400 μM of tetcyclacis. Data represent three replications of extracts from 10 endosperm caps. Error bars indicate standard deviation.

β-mannosidase activity

The activity of β-mannosidase in the endosperm cap followed the same trend as endo-β-mannanase activity. In water-imbibed seeds the activity also increased before radicle protrusion. In GA₄₊₇ the activity was higher (2–6-fold) until 8 d of imbibition and decreased thereafter. Slightly lower levels of β-mannosidase activity were detected in tetcyclacis and in paclobutrazol (Fig. 9B) as compared with the water control.

Tetrazolium stain

Embryos lost their viability when the seeds were imbibed in GA solutions. This was particularly pronounced in the axis, which became brown, confirming that the tissue had died. In the control the axis showed an intense red colour, indicating that the embryos were respiring and alive (Fig. 10A).

Fig. 10.

(A) Tetrazolium-stained coffee embryos from seeds imbibed in water (control) or in 100 μM of GA₄₊₇. After 14 d of imbibition the embryos were isolated and incubated in 0.1% (w/v) of 2,3,5 triphenyltetrazolium chloride for 16 h. (B, C, D) Scanning electron micrographs of the hypocotyl from GA₄₊₇-imbibed seeds at increasing magnification. The embryos were isolated from seeds imbibed for 14 d in 100 μM of GA₄₊₇ solutions. Arrows represent regions with death cells.

Cryo-scanning electron microscopy

Since tetrazolium staining demonstrated that the embryonic axis of GA₄₊₇-imbibed seeds was dead, that particular region was visualized with cryo-SEM. The axis regions contained patches of deteriorated cells (Fig. 10B). Higher magnification revealed that the cells in this region had collapsed and/or had lost cellular compartmentalization (Fig. 10C, D). The dead or dying cells were surrounded by intact cells. The cell contents of the collapsed cells could be observed in the intercellular spaces of the axis cells (Fig. 10D). The groups of dead cells were located in the epidermis, cortex, and in the vascular region. Water- and mannose-imbibed seeds only showed normal turgid cells (results not shown).

Discussion

GAs stimulate and inhibit coffee seed germination

GA₄₊₇ substantially inhibited germination of coffee seeds at a concentration as low as 1 μM. The response to GA₄₊₇ exhibited two sensitivity thresholds: a lower one between 0 and 1 μM and a higher between 10 and 100 μM. This may be caused by a heterogeneous population, consisting of sub-populations of seeds, displaying different sensitivities to the hormone. However, previous dose–response experiments with ABA never indicated any heterogeneity of the seed batch (results not shown). The two steps of inhibition of germination may also be caused by two sites or mechanisms of inhibition with different sensitivities.

When GA-biosynthesis was blocked by tetcyclacis or paclobutrazol, applied GA₄₊₇ stimulated germination up to the optimum of 2 μM after which it became inhibitory. Clearly, the amount of applied GA adds up to the endogenously synthesized hormone. From these data it was estimated that the amount of GAs synthesized in the seed is in the order of a few μM of exogenous GA equivalents. It is not known to what percentage applied GAs are taken up by the coffee seeds. The optimal range of GA concentrations for germination appeared to be very narrow (Fig. 3).

These results indicate that germination of coffee seeds depends on de novo synthesis of GAs. This has been shown for a large number of species, including Arabidopsis and tomato (Karssen et al., 1989; Nambara et al., 1991). However, with current knowledge, coffee is the only species that displays inhibition of germination by GAs at physiological concentrations.

The site of GA action has been proposed to be both in the endosperm and in the embryo (Karssen et al., 1989). In tomato seeds GA induces both embryo growth (Karssen et al. (1989) and endosperm cap weakening (Groot and Karssen, 1987; Groot et al., 1988). In the absence of GA biosynthesis the coffee embryo grew approximately 20%, which was largely due to the swelling of the tissue as a result of hydration that was only completed after 3 d. No increase in pressure potential was measured. Further growth required GAs and was accompanied by an increase in turgor (Fig. 5A). In the endosperm cap the required puncture force decreased at a significantly higher rate in seeds imbibed in GA₄₊₇ than in water. This correlated well with the higher endo-β-mannanase and β-mannosidase activities in GA₄₊₇. The required puncture force decreased until 8 d of imbibition and levelled off thereafter when GA-biosynthesis was blocked by tetcyclacis or paclobutrazol. Also endo-β-mannanase and β-mannosidase activities were considerably lower than in water-imbibed seeds. For radicle protrusion a decrease in the required puncture force below 0.5 N was required. These results indicate that endogenous GAs are required both for embryo cell elongation and endosperm cap weakening during coffee seed germination.

The mechanism of inhibition by supra-optimal GA concentrations

Incubation of seeds up to 14 d in 100 μM of GA₄₊₇ affected embryo viability and larger vacuoles and collapsed cells were observed in the hypocotyl region in the cryo-SEM studies. However, isolated coffee embryos were able to grow on a medium containing 100 μM GA_4+7. Thus, GA itself is not toxic to coffee embryos. Takaki and Dietrich (1980) proposed that mannose, released upon endosperm degradation by endo-β-mannanase and β-mannosidase is involved in the inhibition of radicle protrusion by GA. The cell walls of the coffee seed endosperm are mainly composed of mannans (Wolfrom et al., 1961). Mannose has been shown to inhibit ATP synthesis and normal hexose metabolism (Herold and Lewis, 1977). Since the inhibitory effect of GA on germination leading to embryo death was only observed when the embryo was surrounded by the endosperm, the effect of mannan degradation products on embryo growth was studied. However, the involvement of mannose could not be confirmed: germination was still inhibited when phosphate sources were added together with GA₄₊₇ although it is not clear whether ATP can penetrate unaltered into the embryo cells. Isolated coffee embryos grew normally in a medium containing 10 mM of mannose. In addition, cryo-SEM studies did not even show dead or damaged cells in 100 mM of mannose. Thus, phosphate starvation or other possible effects of mannose are unlikely to be the cause of inhibition of radicle protrusion in coffee seed by supra-optimal GA concentrations. Also other monosaccharides that are likely to increase upon hydrolysis of mannans and galactomannans, including glucose and galactose, had no effect on embryo growth.

These results suggest that supra-optimal GA concentrations release one or more factors from the endosperm that induce cell death in embryonic tissues. It is believed that this is the cause of inhibition and that it occurs very late during the germination process just prior to radicle protrusion. These data show that both embryo growth and endosperm degradation occur in GA largely as in water-imbibed control seeds. Also the protuberance caused by the growing embryo was observed (results not shown). In the presence of GA-biosynthesis inhibitors only the second phase of the endosperm weakening was inhibited as well as endo-β-mannanase activity. Thus, it may also be concluded that the first step of endosperm weakening is independent of the action of GAs.

Cell death is part of a developmental process after a tissue has fulfilled its role and GAs appear to control this process (Fath et al., 2001). In the coffee seed GAs are required to weaken the endosperm and release nutrients from storage cells. Bethke et al. (2001) have hypothesized that hormones regulate cell death in cereal aleurone, which is ultimately induced by reactive oxygen species. These results showed that the embryonic axis turned brown in 100 μM of GA₄₊₇, an indication of oxidative stress and/or the absence of sufficient ‘reducing power’. Thus, what triggers cell death in the endosperm cap during or after its degradation may affect the embryo as well, leading to cell death and, consequently, inhibition of radicle protrusion.

In the coffee seed, exogenous GAs speed up germination-related processes, for example, endosperm cap weakening. It is possible that under these conditions normal cell death of the endosperm occurs too early, with respect to embryo growth. The embryo would then be affected by the damaging components from the endosperm cells. In other words, too much GA disregulates the synchronization of germination processes occurring in the embryo and endosperm.

As yet, it can only speculated about the ecological relevance of GA-inhibited germination. An argument in favour of a possible ecological significance of this phenomenon is that light inhibits germination of coffee seeds (Valio, 1976; Fig. 1). This makes sense in an ecological context since Coffea arabica is originally classified as a shadow plant (Rena and Maestri, 1986). Light induces GA-biosynthesis in seeds (Hilhorst and Karssen, 1992; Toyomasu et al., 1993). To avoid germination under full sunlight coffee seeds may have developed this inhibition mechanism. Yet, it seems unlikely that, under natural conditions, embryonic cell death occurs as a result of high light intensities. However, at lower levels, factors that contribute to cell death may inhibit cellular processes rather than kill the cells.

We thank CAPES (Fundação e Aperfeiçoamento de Pessoal de Nível Superior) for financial support of the studies of EA Amaral da Silva. The seed laboratory at Lavras Federal University, MG, Brazil (UFLA) is acknowledged for handling and shipping the seeds to The Netherlands. We also thank Ms Katja Grolle of the department of Food Science for her technical advice with the material tester. Plant Research International is acknowledged for the use of the psychrometer. Professor W Rademacher of BASF (Germany) is acknowledged for his gift of tetcyclacis and Syngenta (Enkhuizen, The Netherlands) for the gift of paclobutrazol.