Four seed-quality measures in orchids with different pollination systems

Abstract

The number of seeds is an indicator of plant fitness. We compared four quality measures – seed number, abortion, viability and germination. To test as wide a span of seed variability as possible, we cross-pollinated and self-pollinated one nectarless and one nectar-rewarding orchid species, assuming that nectariferous species should be more adapted to geitonogamous selfing than the cheating one and that seed quality should reflect that. Seed number for both species was lowest in selfed fruits. Viability and germination did not show any difference in deceptive Orchis militaris treatments, but the trend was different in rewarding Platanthera bifolia. Seed number and seed abortion correlated well as did viability and germination in vitro. This shows that these two stages are controlled by different mechanisms. Therefore, seed number or seed abortion rate alone cannot be proxies for reproductive output.

Keywords:

• orchid
• seed quality
• seed number
• seed viability
• tetrazolium staining
• germination
• pollination pattern

Introduction

Fruit set and several measures of seed quality have been used in orchid biology as a proxy of plant fitness (Ackerman 1986; Calvo and Horvitz 1990; Tremblay 2011; Roy et al. 2013). Fruit set is an indication that pollination has occurred, but it does not reveal differences in reproductive success because it does not assess the status of the seeds within the fruit. Seed quality has been evaluated by several different measurements. Fruit mass and/or size have been measured, along with seed number or the proportion of seeds with embryos (Sletvold and Ågren 2011; Faast, Facelli, and Austin 2011; Lemay et al. 2015). Other authors have compared embryo viability tests or conducted germination tests (Kauth, Kane, and Vendrame 2011; Meekers and Honnay 2011). Optimal germination conditions must also be considered. To our best knowledge, we do not know of any instance in which all of the available metrics for assessing seed viability have been tested for one or more species simultaneously.

About one-third of all orchids do not offer any reward to pollinators (Dafni 1987; Ackerman 1986; Swartz and Dickson 2009 Tremblay et al. 2005), and there is a distinct difference in pollinator behaviour between rewarding and non-rewarding species. Foraging pollinators typically visit each flower on the inflorescences of rewarding species and do not leave the inflorescence until all the flowers are visited. In contrast, the inflorescences of non-rewarding species are often quickly abandoned by deceived pollinators (Johnsson and Nilsson 1999). This suggests that deception increases allogamy whereas rewarding species should experience geitonogamy relatively often. Several authors have shown that deception does promote cross-pollination (Johnsson and Nilsson 1999; Jersáková, Johnson, and Kindlmann 2006; Jersàková and Johnson 2006). Ackerman, Cuevas and Hof (2011) concluded that populations of deceptive species are more variable than rewarding species in traits associated with pollinator attraction. However, there is evidence that the reward system does not always promote geitonogamy (Johnson and Nilsson 1999). Rewarding and deceptive species differ in their levels of natural fruit set (Neiland and Wilcock 1998; Jersáková, Johnson, and Kindlmann 2006) and inbreeding depression (measured in percentage of aborted seed) is higher in rewarding species (Sletvold et al. 2012). Cross-pollination is thought to enhance seed quality compared with geitonogamy (de Jong, Waser, and Klinkhamer 1993).

We aimed to understand how different quality measures (seed number, seed abortion, seed viability by tetrazolium test and germination in vitro) reveal actual seed quality.

Material and methods

We chose two orchid species with different pollination systems to enable assessment of a range of seed qualities, namely the food deceptive Orchis militaris and the nectar rewarding Platanthera chlorantha. Total seed numbers were counted in the capsules, the ratio of empty seeds to seeds with embryos was calculated, the seeds were stained to evaluate viability, and finally, the seeds were germinated in vitro.

Orchis militaris L. and Platanthera chlorantha (Custer) Rchb. are terrestrial tuberous orchids, widely distributed across Eurasia (Hulten and Fries 1986). They grow in open scrub and grassland habitats, typically on soils with high pH and generally co-occur. Orchis militaris is a non-rewarding species with sweet-scented flowers that do not contain nectar (Bell et al. 2009). Seed set in O. militaris is reported to be poor since only 2–28% of the flowers produce capsules (Farrell 1985). Platanthera chlorantha is a rewarding species with greenish-white flowers that produce a vanilla scent, and each flower has a long spur that contains nectar (Stpiczyńska 2003; Bell et al. 2009). A fruit set of approximately 60% has been measured (Nilsson 1983). Various Lepidoptera Linnaeus, 1758 and Hymenoptera Linnaeus, 1758 species (Farrell 1985; Steen 2012) pollinate O. militaris. Platanthera chlorantha is pollinated mostly by the moths from the Sphingidae Latreille, 1802 and Noctuidae Laterille, 1809 (Nilsson 1978; Steen 2012).

Field and laboratory methods

The studied population of Orchis militaris was situated on an abandoned pasture, overgrown with junipers, on the western Estonian island Muhu (58°38'55.5252"N, 23°19'8.5404"E), and consists of approximately 200 individuals. The Platanthera chlorantha population was located in a semi-dry meadow in southern Estonia (58°20'44.16"N, 26°53'8.22"E), and encompasses approximately 100 individuals.

In 2011, we randomly chose 64 O. militaris and 82 P. chlorantha sprouting generative individuals in two populations. Plants assigned to the artificial pollination group were enclosed with fine nylon-mesh netting before anthesis of the flowers.

The three following treatments were performed: (1) cross-pollination, (2) geitonogamous self-pollination and, as a control, (3) natural pollination. Pollen donors for cross-pollination were selected from plants at least 10 m away. One pollinium per flower was used in the two first treatments. We collected mature capsules after approximately 2 months and assessed four indicators of seed production and seed viability: number of seeds per capsule, the ratio of seeds with and without embryos in each fruit, seed viability and seed germination in vitro.

The number of seeds per fruit was counted under a light microscope after the contents of each fruit were placed in a Petri dish. Each seed was also scored for the presence or absence of an embryo.

Pre-treatment for viability staining and the germination experiment included soaking in hypochlorite. The tetrazolium staining (TZ) test was used to estimate potential viability (Lakon 1949). Tetrazolium indicates the activity of enzymes from the dehydrogenase group, which is responsible for the reduction process in living tissue (Singh 1981). A share from each fruit was used to determine viability and, since living tissue in embryo stains red, we counted only fully red embryos as viable, not partially coloured or pale, reddish embryos.

For asymbiotic germination, we used the commercial (PhytoTechnology Laboratories, Overland Park, KS, USA) germination medium BM-1 that has been successfully used on other orchid species by several authors (Van Waes 1984; Dutra et al. 2008; Kane 2009).

Seeds from each capsule were planted in three Petri dishes, which served as replicates. Germination was attempted in complete darkness with stratification at 2°C for 3 months before continuing for another 6 months at 20 ± 2°C. We considered seeds to have germinated when they formed a well-developed protocorm, not considering the seeds that were just swollen.

Data analysis

We report below mean ± standard error. A logistic regression model was used to estimate the odds ratios for aborted versus normal seeds. A logit transformation log(p/(1–p)), where p is the proportion of aborted seeds, was used before calculating the odds ratios (p_i/1–p_i)/(p_j/(1–p_j) between the treatments i and j.

The germination percentage and the viability data were transformed using the root mean square (sqrt) and arcsine (asin) functions in the R statistical package. The transformed data were analysed using a one-factor fixed effects analysis of variance (ANOVA) model to compare differences in the pollination treatments and the differences between species. Tukey test (honest significant difference) for unequal sample size was used to test for post hoc comparisons. To control for interference from seed-quality measures, we calculated Pearson’s correlation for each pollination treatment by species. All calculations were carried out using statistical package R 3.1.1 (R Core Team, 2014).

Results

Seed number

The naturally pollinated fruits of Orchis militaris produced more seeds per capsule than Platanthera chlorantha (Figure 1). The number of seeds in self-pollinated versus naturally pollinated fruits differed significantly for O. militaris (ANOVA: F_2,62 = 4.185, p = 0.0197), with naturally pollinated fruits containing more seeds (n = 6879 ± 1064). For the rewarding P. chlorantha, the number of seeds per fruit varied significantly (ANOVA: F_2,79 = 31.011 p = 0.0001), and was highest in naturally pollinated fruits (n = 2006 ± 142).

Figure 1. Pollination effect on seed quality on deceptive Orchis militaris (left) and on rewarding Platanthera chlorantha (right): number of seeds per capsule, abortion measured as relation empty/normal seeds, viability by a tetrazolium staining test, germination in vitro. Error bars are SEs (see text) the level of statistical significance of the effect of pollination treatment for each quality indicator (ns, not significant *p < 0.05 **p < 0.01 ***p < 0.001). Bars with different letters [(a), (b) and (c)] are significantly different (p < 0.05) by a Tukey honest significant difference test for unequal sample sizes.

In both species, fruits produced by cross-pollination (allogamy or geitonogamy) contained higher numbers of seeds per fruit than self-pollinated fruits.

Seeds with and without embryos

The percentage of aborted O. militaris seeds was similar in the control and self-pollinated fruits (45.6% and 45.4%, respectively) and was lowest (38%) in out-crossed fruits (Figure 1). In P. chlorantha controls, 23.6% of the seeds had no embryos, compared with 46.2% of self-pollinated and 50.1% of out-crossed fruits. In deceptive O. militaris, the abortion level was quite similar across pollination treatments. Significant differences were not found, though self-pollinated seeds showed slightly lower quality. For P. chlorantha, the empty : full ratio values differed significantly between all treatments and were especially high for self-pollination.

Viability

There was a small difference in the number of viable seeds in the three treatments of O. militaris (ANOVA: F_2,62 = 3.45, p = 0.034), with cross-pollinated seeds showing higher viability than those exposed to the other two treatments. In P. chlorantha, the differences between the treatments were not significant (ANOVA: F_2,79 = 2.6 p = 0.07), with the lowest viability in the self-pollinated flowers (Figure 1).

Germination

The germination percentages differed in the three treatments of O. militaris (ANOVA: F_2,189 = 3.623, p < 0.03), with a significantly lower germination in control flowers (Figure. 1). In P. chlorantha, there were also significant differences (ANOVA: F_2,234 = 25.4 p = 0.0002), and naturally pollinated flowers had seeds with a higher germinating percentage. The differences in allogamous and geitonogamous treatments were non-significant. Figure 1 shows that the trends of the control – crossed – self-pollinated seeds of the two species run counter to each other.

Correlations between the indicators

There was a negative correlation between seed number and abortion in all O. militaris pollination treatments (Table 1), and the only positive correlation was between seed germination and viability in the controls. In the rewarding P. chlorantha (Table 2), there was a negative correlation for all treatments and a positive correlation between germination and viability in all three pollination treatments.

Table 1. Pearson’s correlation coefficients and their statistical significance between seed quality measures: (1) seed number (2) abortion (3) viability (4) germination) in deceptive Orchis militaris for three pollination treatments.

Control	(1)	(2)	(3)	(4)
(1) Seed number	1.000
(2) Abortion	–0.630**	1.000
(3) Viability	–0.044	–0.083	1.000
(4) Germination	–0.117	–0.141	0.545**	1.000
Cross-pollination
(1) Seed number	1.000
(2) Abortion	–0.607***	1.000
(3) Viability	-0.099	–0.069	1.000
(4) Germination	0.145	–0.118	–0.098	1.000
Self-pollination
(1) Seed number	1.000
(2) Abortion	–0.446*	1.000
(3) Viability	0.125	–0.379	1.000
(4) Germination	–0.017	–0.225	0.422	1.000

^* p < 0.05

^** p < 0.01

^*** p < 0.001.

Table 2. Pearson's correlation coefficients and their statistical significance between seed quality measures in rewarding Platanthera chlorantha for three pollination treatments.

Control	(1)	(2)	(3)	(4)
(1) Seed number	1.000
(2) Abortion	–0.407***	1.000
(3) Viability	0.292	–0.138	1.000
(4) Germination	0.088	–0.240	0.397***	1.000
Cross-pollination
(1) Seed number	1.000
(2) Abortion	–0.544*	1.000
(3) Viability	–0.092	0.050	1.000
(4) Germination	–0.287	–0.075	0.540*	1.000
Self-pollination
(1) Seed number	1.000
(2) Abortion	0.576**	1.000
(3) Viability	0.088	0.231	1.000
(4) Germination	0.390	0.253	0.983***	1.000

^* p < 0.05

^** p < 0.01

^*** p < 0.001.

Every seed-quality indicator we tested was unique (Tables 1 and 2), but there were logical correlations between seed number and abortion, viability and germination (with the exception of O. militaris, which revealed a clear correlation between only abortion and seed number). Seed number did not correlate with viability or germination in either species.

Discussion

In our experiment, the self-pollination of flowers resulted in lower seed numbers and an increased percentage of seeds without embryos in both species, but the trend was significant only for the rewarding P. chlorantha. The seed numbers in the fruits that were artificially pollinated were lower than natural ones in both species. It is suggested that rewarding species receive multiple pollination events or pollen from multiple donors, and so are more allogamous, as is confirmed by the pattern in P. chlorantha seed counts. However, O. militaris showed the same pattern when only seed number was counted. In rewarding Platanthera, abortion in the controls was relatively low compared with the artificial treatments. In deceptive O. militaris, seed abortion was at the same level in all treatments, including naturally pollinated ones. This also suggests that rewarding species produce high-quality seeds because of a relatively high visitation rate. The pollinators of rewarding P. chlorantha visit 9.6 flowers on average, and an average visit lasts for 38.0 seconds (Steen 2012). Deceptive species generally have low visitation frequencies (Neiland and Wilcock 1998). These species seem to have evolved to specialize in one pollination chance instead of maximizing pollinator attraction, they maximize the seed production that results from every casual encounter (Pérez-Hérnandez et al. 2011). The trends for rewarding P. chlorantha remained the same in viability and germination tests, but deceptive O. militaris showed somewhat opposing results for the last two seed-quality measures. Naturally pollinated O. militaris seeds had the lowest germination in this study, indicating that there might be a problem with the suitability of the pollen (potentially due to outbreeding) or some other behavioural factor. This goes unnoticed when we only measure capsules or count seeds.

The TZ and in vitro germination tests have similar potentials to measure viability and potency in progeny, as shown in Tables 1 and 2. Some authors have tried to find correlations between TZ and germination in vitro without success (Shoushtari et al. 1994; Rasmussen and Whigham 1993) but others have found positive relationships between these factors (Van Waes and Debergh 1986; Pant, Purohit, and Lal 1999; Lemay et al. 2015). Because TZ staining is much quicker and a considerably cheaper indicator of seed quality than germination in vitro, we propose using it in combination with a seed count indicator. Using just one of the indicators may lead to the wrong conclusions in some species, such as in O. militaris, where TZ staining and germination percentages are not in accordance with seed number and abortion.

Our results coincide largely with the results of several authors (Nilsson 1983; Johnsson and Nilsson 1999; Vallius, Arminen, and Salonen 2006; Sletvold et al. 2012) who have studied mainly nectariferous orchids (Platanthera and Gymnadenia species) using fruit measures, seed numbers, seed abortion and germination as separate indicators of reproductive success. Sletvold et al. (2012) demonstrated rather contrasting results on rewarding Gymnadenia conopsea (L.) R. Br., where estimates of inbreeding and outbreeding depressions were quite different for germination and for seed production. Differences in measures may also arise from the fact that in many terrestrial orchids, immature seeds with incomplete embryos germinate more readily in vitro than mature ones (Arditti 1992; Yamazaki and Miyoshi 2006). When germinating mature seeds, one must address different levels of seed dormancy (Rasmussen 1995; Baskin and Baskin 1998; Whigham et al. 2006), and several researchers have observed low levels of germination, especially during the first 6–9 months (Vujanovic et al. 2000).

Every seed-quality indicator we tested was unique, and there are good correlations between seed number and abortion, viability and germination. We can see that when only one or two of these measures – the fruit mass, seed number and the seed with or without the embryo – are taken into account, fitness-affecting factors may easily be missed. Different seed-quality indicators are useful for obtaining the maximum information about a study species. Rewarding and deceptive species show different trends in progeny through the levels of different seed-quality measures. Seed number did not correlate with viability or germination in either species, so this alone should not be used for fitness or depression or other index calculations. Accordingly, for fitness and depression calculations more than one uncorrelated measure should be used. As we only tested two species in this study, more species should be analysed in the future to see how frequent such differences are between species with different biology.

Conclusion

In unrewarding species, simple seed counts may not give sufficient information about the actual progeny viability. Artificial treatments with a single pollination event suggest that for both rewarding and non-rewarding species, the pollen origin influences the seed number and abortion in a similar way, but germination may exhibit more complicated patterns. The number of pollination events or the amount of pollen from multiparental donors can also make a difference in actual reproduction.

Notes on contributors

Mirjam Metsare is a third-year doctoral student in the Department of Botany, Institute of Agricultural and Environmental Sciences. She has been working on orchid seed germination. Contribution: most of the laboratory work and manuscript writing of this research.

Aigi Ilves is a fourth-year doctoral student in the Department of Botany, Institute of Agricultural and Environmental Sciences. Her work includes orchid pollination and genetics. Contribution: fieldwork during this research.

Marina Haldna works as a Research Scientist in the Institute of Agricultural and Environmental Sciences. Contribution: main theme is statistical analyses.

Tiiu Kull is a Professor in the Department of Botany, Institute of Agricultural and Environmental Sciences. She has long studied terrestrial orchids. Contribution: contributed to this research with the idea and project coordination.

Kadri Tali is a Senior Researcher in the Department of Botany, Institute of Agricultural and Environmental Sciences, studying orchid biology and germination. Contribution: financed and coordinated the project and supervised manuscript preparations.

Acknowledgements

This research was supported by Estonian SF grant no 8584 and by institutional research funding IUT21-1 of the Estonian Ministry of Education and Research. We thank the herbarium of vascular plants and mosses (TAA) for support. We are very grateful to Dr Marc-Andre Selosse and the anonymous reviewers for their detailed and very helpful comments, we also acknowledge the staff of the TORC’15 International Conference on Temperate Orchids Research & Conservation and “Sails-for-Science” foundation.