Biological Journal of the Linncan Socicly (1993), 49: 239-248. With 3 figures Interspecfic size and number variation in pollen grains and seeds WILLIAM D. J. KIRK Department of Biological Sciences, Keele Universily, Keele, Slafordshire ST5 5 B G Rcceiued 31 March 1992, accepled for publication I4 M a y 1992 An interspecific correlation between pollen grain size and seed size is demonstrated by means of the phylogenetic regression, which allows for phylogenetic bias. The correlation was not explained by plant size, mass of DNA per cell, style length or breeding system, although the first three of these factors all correlated with both pollen size and seed size. Two interpretations, involving pollen competition and flower size, are discussed. There is also an interspecific correlation between pollen grain number per flower and ovule number per flower. Some consequences of these correlations for the interpretation of pollen-ovule ratios and sex allocation strategies are considered. ratio - flower ADDITIONAL KEY WORDS:-Pollen-ovule allocation. CONTENTS Introduction . . . Materials and methods Statistical analysis Plants . . . Flowers . . . Seeds . . . Compositae . . Cruciferae . . Results . . . . Discussion . . , Acknowledgements . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - DNA - style - competition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . - sex 239 240 240 24 1 24 1 242 242 242 242 245 246 247 INTRODUCTION The mean seed size of a species has many repercussions for the reproductive ecology of plants (Harper, Love11 & Moore, 1970). Seed size is linked with habitat characteristics such as soil disturbance, shade and drought (Salisbury, 1942; Baker, 1972). It is also linked with dispersal ability and survival in seed banks (Grime, Hodgson & Hunt, 1988). Taxonomic constraints appear to have reduced intra-familial variation (Hodgson & Mackey, 1986), while developmental constraints may be responsible for correlations of seed size with fruit size, flower size and plant size (Primack, 1987; Silvertown, 1989). Variation in the mean pollen grain size of a species is less well understood. The main factor historically linked with pollen size is style length, but the interpretation remains controversial with evidence for and against a causative 239 0024-4066/93/070239+ 10 ,608.00/0 0 1993 The Linnean Society of London 240 W. D. J. KIRK link (Darwin, 1877; Cruden & Lyon, 1985; Williams & Rouse, 1990; Cruzan, 1990), although differences between binucleate and trinucleate pollen may have caused some confusion (Bertin, 1988). Cruden & Lyon (1985) showed that pollen size could instead be reflecting stigma depth, the distance a pollen tube has to grow to reach exogenous resources. If so, the assumption that pollinator size determines style length and hence pollen size (Taylor & Levin, 1975; Lee, 1978) is considerably weakened. A variety of other influences on inter-specific differences in pollen size have been suggested. They include: aerodynamic constraints for wind-pollinated and buzz-pollinated species (Whitehead, 1969; Muller, 1979), polyploidy (Levin, 1968; Muller, 1979), breeding system (Pandey, 197 l ) , longevity (Cruzan, 1990), the pollen population effect (Cruden & Miller-Ward, 1981), pollinator dietary requirements (Baker & Baker, 1979) and competition between pollen grains (Willson & Burley, 1983). Seeds and pollen can be considered in the context of evolutionary trade-offs between size and number (Smith & Fretwell, 1974; Lloyd, 1987, 1988). These trade-offs cover an enormous size range for plant propagules (Haig & Westoby, 1991). In the British flora, the largest seeds are at least 100000 times as heavy as the smallest and the largest pollen grains have at least 1000 times the volume of the smallest. However, investigations of pollen-ovule number ratios (P/Os) in relation to pollination mechanism and breeding system have generally not taken into account the direct effect of a size-versus-number trade-off on PI0 variation (Cruden, 1977; Cruden & Miller-Ward, 1981). Charnov (1982) gives an equation linking PI0 with pollen size and seed size, assuming size-versus-number trade-offs and ovule number equal to seed number: 'J log ( P / O )= log (1- + logC*-logC, where r is the proportion of reproductive resource allocation given to pollen, C, is the cost of a seed and C,is the cost of a pollen grain. Charnov (1982) pointed out that, assuming seed size shows no systematic interspecific variation with pollen size, this equation may be sufficient to explain negative correlations between P I 0 and pollen size (Cruden & Miller-Ward, 1981) and positive correlations between P/O and seed size (Shaanker & Ganeshaiah, 1984; Preston, 1986). Queller ( 1984) developed Charnov's equation to show that correlations between P/Os and four pollination and breeding system factors can be interpreted in terms of evolutionary models of optimal allocation of resources to male and female functions, rather than in terms of efficiency (see Cruden, 1977; Cruden & Miller-Ward, 1981). In this paper, I show that seed size does show an interspecific correlation with pollen size and I investigate several interpretations of the relationship. An understanding of the links between seed size, seed number, pollen size and pollen number is relevant to studies of sex allocation strategies in hermaphrodite plants. MATERIALS AND METHODS Statistical analysis The phylogenetic regression of Grafen (1989) was used because it allows for phylogenetic bias in cross-species data (Page1 & Harvey, 1988) and permits one POLLEN AND SEEDS 24 I to control for other variables. The method needs a ‘working phylogeny’. For this analysis, the phylogeny was constructed from a published classification, with each of the named taxa in the classification (the genera, families etc.) forming the nodes in the phylogeny. The distances between nodes (path segment lengths) were assigned by the ‘taxonomic levels method’. Nodes at successive taxonomic levels were assigned the numbers 1, 2, 3 etc. and path segment lengths calculated from the differences in these values. The distances were power transformed and used to provide a set of variances and covariances for the regression (see Grafen, 1989). Non-independence between species is divided into that which is and that which is not due to associations represented in the working phylogeny. The former is dealt with by developing the standard regression and the latter by using each node of the phylogeny as an independent data point. The error degrees of freedom depend on the number of nodes in the phylogeny and not the number of species involved. Generalized least squares lines of fit are plotted. Values o f ? are quoted to show how much variance is accounted for, but the implied P-values are not reliable tests of the regression. Results were computed by means of a program (version 1.03) in GLIM, supplied by A. Grafen (personal communication, 1991). All sizes and masses were log,,, transformed. Plants The analyses of English species were restricted to angiosperm species listed in the flora of the central English county of Warwickshire (Cadbury, Hawkes & Readett, 1971). As part of further studies not reported here, species were excluded that were escapes, introduced, planted, naturalized, casuals, relics of cultivation or non-native, according to Cadbury et al. (1971), or that were water-pollinated or exclusively apomictic, according to Faegri & van der Pijl (1979) or Proctor & Yeo (1973). The full analysis included 610 species in 81 families. The number of species for which data could be obtained are given in parentheses for each variable below. The classification of genera given by Cadbury et al. (1971) and of tribe, sub-family, family, order and sub-class given by Clapham, Tutin & Moore (1987) was used in the phylogenetic regression. Plant size (584 spp.) was obtained from the mid-range normal fully grown plant height or stem length given by Clapham et al. (1987). Floating, creeping or climbing species were not directly comparable and were omitted. DNA content (191 spp.) was obtained from the mass of 2C DNA per nucleus in root-tip cells given by Grime et al. (1988). Flowers Pollen grain size (559 spp.) was obtained from the length of the long axis of acetolysed grains in silicone oil recorded by Andrew ( 1984). Volumes could not be used in the analysis because the lengths of the short axes were not available for calculating the volumes of the non-spherical grains. The mid-range size was used for species with more than one size of pollen. Tetrad sizes were divided by two to obtain grain sizes. Seven species omitted by Andrew were measured from specimens in the Department of Geography, Keele University and the Department of Botany, University of Cambridge. Species were categorised by pollination mechanism according to Faegri & van der Pijl (1979), Proctor & Yeo W.D. J. KIRK 242 (1973) and Clapham et al. (1987) into wind-pollinated (167 spp.) and insectpollinated (all others) (443 spp.). Seeds Seed size classes (403 spp.) were taken from the data of Grime et al. (1988) for the air-dried mass of a single seed, achene or other indehiscent germinule: 1 ( < 0.20 mg), 2 (0.21-0.50 mg), 3 (0.51-1.00 mg), 4 (1.01-2.00 mg), 5 (2.01-10.00 mg), 6 ( > 10 mg). These classes reduce the effect ofoutliers and are roughly logarithmic. Compositae Pollen size (60 spp.), plant size (61 spp.) and mass of DNA (22 spp.) were obtained as above. Pollen volume was calculated from pollen length, d, with the , the grains are spherical or nearly so. Seed mass (29 spp.) formula ( n / 6 ) d 3 since was taken from the 'germinule mass' (air-dried) given by Grime et al. (1988). Style length (63 spp.), from the apex of the ovary to the base of the stigma, was measured from the botanical drawings of Ross-Craig ( 1960-1 963). Breeding systems (24 spp.) were divided according to Grime el al. (1988) into three classes: out-crossing, both selfing and out-crossing, and predominantly selfing, scored 1, 2 and 3, respectively. Cruciferae The data of Preston (1986) for 66 Californian taxa were used. Numbers of ovules and seeds were log,, transformed. Phylogenetic regressions were performed as above, with the classification into genera and species used by Preston ( 1986). RESULTS The regression of pollen size on seed size class was significant (F,,,35= 38.0, P < 0.001, N = 375, ?(3,3, = (+)0.10) (Fig. 1). A similar result held within just 0 0 0 '" 6 1 2 3 4 6 6 Seed size class Figure 1. Pollen grain length plotted against seed size class, log,,T= 0.027 log,,X+ 1.334. Key: dicots ( 0 )monocots ; Points have been offset for seed size class to reduce overlap. (v). POLLEN AND SEEDS 243 1001 A 1 0.1 ':'obc8c\P81.0 W ' 10.0 DNAW Figure 2. A, Pollen grain length plotted against the mass of 2C DNA per nucleus in root-tip cells, log,,T = 0.071 log,,X+ 1.378. B, Seed size class plotted against the m a s of 2C DNA, log,,?'= I .016 logI,X+2.73l. Key: dicots (a);monocots (V). Points have been offset for seed size class to reduce overlap. the wind-pollinated species ( F l , 3 3 = 24.0, P < 0.001, N = 107, = (+)0.20) and just the insect-pollinated species (F,,%= 21.0, N = 268, P < 0.001, ?(2ss) = (+)0.08). Developmental constraints might account for this correlation if plant size constrains seed size and pollen size, perhaps through a constraint on flower size. The regression of seed size class on plant size was significant (F,,,,,= 22.2, P < 0.001, N = 385, r2 383) = (+)0.06), confirming the findings of others (see Silvertown, 1989), but the regression of pollen size on plant size was only weakly = 6.0, P = 0.02, N = 537, ?(535) = (+)0.01). The regression of significant (F1,183 pollen size on seed size class was virtually unchanged by controlling for plant size (F,,,37= 35.4, P < 0.001, N = 359, ?(356) = (+)omlo). The amount of DNA per cell might account for the correlation. Polyploid forms of plants have larger pollen grains (Levin, 1968; Muller, 1979), suggesting that the amount of DNA constrains pollen size. The regression of pollen size on = 8.6, P = 0.004, N = 181, ?(179) = mass of DNA per cell was significant (+)0.05) (Fig. 2A). Plants with early spring growth have more DNA per cell (Grime, Shacklock & Band, 1985; Thompson, 1990), and such plants might also have larger seeds with more endosperm to allow them to grow faster and sooner (Stanton, 1984). Polyploid seeds also weigh more (Salisbury, 1942). The regression of seed size class on mass of DNA per cell was also significant ( F I , 6 1 = 12.4, P < 0.001, N = 180, ?(178) = (+)0.07) (Fig. 2B). However, the W. D.J. KIRK 244 60 O O O L A f fBB 0 0. / 0 v v a 10 000 -@ a d EiO00 $ 6000& // -0 ---' 0 ow 0 ,a 1 I I 10.0 1.0 0.1 Seed mass (mg) 60 000 f v B c ( g 10 000 Q c ( & 6000 0 Style length (mm) 0 Style length (mm) Figure 3. Species in the Compositae. A, Pollen grain volume plotted against seed mass, log,,Y = 0.349 logIJ+4.22l. B, Pollen grain volume plotted against style length, log,,Y = 0.418 log,,X+3.811. C, Seed mass plotted against style length, log,,Y = 1.020 log,,X- 1.168. Key: ; (V). Asteroideae ( 0 ) Cichorioideae regression of pollen size on seed size class remained significant when controlling = (+)0.10) or for mass of DNA (F1,77= 14.5, P < 0.001, N = 171, controlling for plant size and mass of DNA (Fl,75 = 12.1, P < 0.001, N = 166, '(162) = (+)o'09)' Some further factors were examined for the family Compositae. The regression of pollen volume on seed mass was just significant (Fl,,o= 5.7, P = 0.04, N = 27, ?(25) = (+)0.25) (Fig. 3A). Style length often correlates with pollen size (see Introduction). The regression POLLEN AND SEEDS 245 of pollen volume on style length was significant for Compositae (Fl,17 = 6.2, P = 0.02, N = 60, ?,58, = (+)0.10) (Fig. 3B), as also was the regression of seed = 6.5, P = 0.03, N = 29, ?(27) = (+)0.22) (Fig. 3C). mass on style length (Fl,ll However, the regression of pollen volume on seed mass was still significant when = 10.6, P = 0.009, N = 27, ?(24) = (+)0.25). controlling for style length (F,,lo The relationship between breeding system and pollen and seed sizes is of particular interest because of the correlation between breeding system and PI0 (see Introduction). The regression of pollen volume on breeding system was not significant (F,,9= 3.13, P = 0.1 1, N = 23, r2(21)= (+)0.03) and the regression of seed mass on breeding system was not significant (Fl,l,= 0.34, P = 0.57, N = 24, 8(22) = (+)0.01). However, the available breeding system data are rather imprecise. The relationship between number of pollen grains per flower and number of ovules per flower was tested with the data of Preston (1986) for 66 crucifer taxa. Standard regressions, which allow comparison with Preston’s analysis, were highly significant for 37 self-compatible taxa ( t ( 3 5 )= 5.5, P < 0.001, ?,, = (+)0.47) and for 29 self-incompatible taxa = 7.3, P < 0.001, ?(27) = (+)0.66). Phylogenetic regressions of pollen grains per flower on ovules per = 11.8, P = 0.006, ?(35) = flower were also significant for these two groups (Fl,lo (+)0.29 and F,,6= 62.4, P < 0,001, ?(2,) = (+)0.65 respectively). DISCUSSION Pollen size and seed size are positively correlated over a wide range of species. Although plant size, mass of DNA and style length correlate with both of them, these factors do not appear to account for much of the pollen-seed size correlation. However, caution is needed in interpreting the role of a factor that is not significant or that does not account for a correlation, since this could be just the result of few or imprecise data for that factor. This is most likely for the breeding system data. Better experimental data on the frequency of self-fertilization in plants would be useful. Willson & Burley (1983) have argued that where pollen competition is low, there should be minimal investment per pollen grain, but where pollen competition is high, pollen should be investment rich. There is some experimental evidence that larger pollen grains are more competitive. Larger grains can have higher rates of tube elongation (Lord & Eckard, 1984; Williams & Rouse, 1990) and higher rates of ‘post-fertilization siring ability’ (Cruzan, 1990). Such post-fertilization zygotic advantages might even be reflected in germination rates and seedling growth. Correlations between gametophytic and sporophytic qualities are already known (Mulcahy, 1983). This suggests a possible interpretation of the pollen-seed size correlation. For any particular male/female allocation of resources, if seed number per flower decreases as a result of selection for larger seeds, then there will be more competition between pollen grains. The same number of pollen grains would be competing to be the parents of fewer seeds. If larger grains are more competitive, this could favour larger pollen when seeds are larger. Unfortunately, there is a paucity of data on paternal performance (Bertin, 1988) and it is not possible to test this hypothesis with existing data. 246 W. D. J . KIRK An alternative interpretation is that seed size and pollen size are constrained by flower size. Although there was no evidence for a constraint imposed by plant size, direct measures of flower size would be a much better test. Several could be tried. Style length is a good measure of floret size in the Compositae, but this size measure did not account for the pollen-seed size correlation within the Compositae. The number of pollen grains per flower (P) and the number of ovules per flower (0) are highly correlated in the Cruciferae. The same trend is apparent across ten families in the data of Cruden & Miller-Ward (1981). This number correlation is likely to be mainly accounted for by the variation in total male/ female resource allocation per flower, rather than by correlated size-versusnumber trade-offs in pollen and seeds. More comprehensive data could elucidate to what extent either or both of these independent trade-offs are involved. Increased reproductive resource allocation per flower appears to be manifested interspecifically by an increased number of pollen grains rather than by an increase in their size (Cruden & Miller-Ward, 1981). The same may be true for seeds. However, if the sizes of pollen and seeds were both increased by higher resource allocation, this could explain the pollen-seed size correlation. Developmental constraints might do this, as proposed for some other size correlations (see Introduction). An adaptive interpretation would need to explain why the optimum seed size and pollen size for a species would be higher when more resources are available per flower. Pollen-seed correlations can affect the interpretation of cross-species data. A strong pollen-seed number correlation would give a high covariance of logP and logo, which would considerably reduce the variance in log(PI0). If this covariance arises from variation in total resource allocation, the reduced variance of log(P/O) would expose variance from other sources, such as maleversus-female or size-versus-number trade-offs. The variance from size-versusnumber trade-offs might also be reduced a little by the pollen-seed size correlation, since pollen and seed sizes do not vary independently. Reduction of a large total resource allocation variance and consequent exposure of a maleversus-female trade-off variance probably accounts for Preston’s ( 1986) finding that log(P/O) was a better indicator of breeding system than logP or log0 alone. Queller (1984) considered that the data of Cruden & Miller-Ward (1981) could not be explained by a ‘simple trade-off between pollen size and number’ because the correlation between logP and log(pol1en size) was not significant, while that between log(P/O) and log(pol1en size) was. My argument above suggests that in surveys covering a wide range of total resource allocation per flower, use of log(P/O) instead of logP could be more, rather than less, likely to expose a sizeversus-number trade-off for pollen. Comprehensive data for the size and number of pollen grains and seeds for a wide range of species would allow these ideas to be tested further. ACKNOWLEDGEMENTS Facilities were provided by: the Department of Biological Sciences and the Department of Geography, Keele University; and the Department of Botany, University of Cambridge. I thank the following for help and advice: POLLEN AND SEEDS 247 S. A. Corbet, A. Grafen, P. W. Jones, S. Peglar, A. Polwart, A. S. Pullin and P. A. Thomas. I am grateful for the comments and suggestions of the referees. REFERENCES Andrew R. 1984. A practical guide to the British Jora. Quaternary Research Association Technical Guide No. I. Cambridge: Quaternary Research Association. Baker HC.1972. Seed weight in relation to environmental conditions in California. Ecology 53: 997-1010. Baker HC, Baker I. 1979. Starch in angiosperm pollen grains and its evolutionary significance. Amm'can journal of Botany 66: 591-600. Be& RI. 1988. Paternity in plants. In: Lovett Doust JL, Lovett Doust, LL, eds. Plant reproductive ecology. Pattern and strategies. New York: Oxford University Press, 30-59. Cadbury DA, Hnwkts JC,R e d c t t RC. 1971. A compuler-mappcdJora. A stu& of the counfy of Warwickshire. London: Academic Press. Charnov EL. 1982. The theory ofsex allocation. Princeton: Princeton University Press. Clapham AR, Tuth TC, Moore DM. 1987. Flora of the British Isles, 3rd edn. Cambridge: Cambridge University Press. Cruden RW. 1977. Pollen-ovule ratios: a conservative indicator of breeding systems in flowering plants. Evolution 31: 3 2 4 6 . Cruden RW, Lyon DL. 1985. Correlations among stigma depth, style length and pollen grain size. Do they reflect function or phylogeny? Botanical Gazette 146: 143-149. Cruden RW, Miller-Wud S. 1981. Pollen-ovule ratio, pollen size, and the ratio of stigmatic area to the pollen-bearing area of the pollinator: an hypothesis. Evolution 35: 964-974. Cruzan MB. 1990. Variation in pollen size, fertilization ability, and post-fertilization siring ability in Erythronium grandijorum. Evolution 44: 843-856. Duwip C. 1877. The da&enl forms ofJowers on plants of the same species. London: Murray. Fntgri K,Pijl L. v.p der. 1979. The principles ofpollination ecology, 3rd edn. Oxford: Pergamon Press. & d e n A. 1989. The phylogenetic regression. Philosophical Transactions of the Royal Sociefy, London, B 326: 119-157. Orimc JP, H+um JC, Hunt R. 1988. Comparative plant ecology. A functional approach to common British species. London: Unwin Hyman. Grime JP, Shacklock JML, Bmnd SR. 1985. Nuclear DNA contents, shoot phenology and species coexistence in a limestone grassland community. New Phytologist 10th 435445. Haig D, Wcstoby M. 1991. Seed size, pollination costs and angiosperm success. Evolutionary Ecology 5: 231-247. Huper JL, Lovell PH, Moore KC. 1970. The shapes and sizes of seeds. Annual Review of Ecology and Systematics 1: 327-356. Hodgson JC, h c k e y JML. 1986. The ecological specialization of dicotyledonous families within a local flora: some factors constraining optimization of seed size and their possible evolutionary significance. New Phytologist 101: 497-5 15. Lee 8. 1978. A factor analysis study of the functional significance of angiosperm pollen. Systematic Botany 3: 1-19. Levin DA. 1968. The genome constituents of eastern North American Phlox amphiploids. Evolution 22: 612632. Lloyd DC. 1987. Selection of offspring size at independence and other size-versus-number strategies. American Naturalist 129: 800-81 7 . Lloyd DC. 1988. A general principle for the allocation of limited resources. Evolutionary Ecology 2: 175-187. Lord EM, E M KJ. 1984. Incompatibility between the dimorphic flowers of Collomia grandijora, a cleistogamous species. Scince 223, 695496. M d d y DL. 1983. Models of pollen tube competition in Geranium maculatum. In: Real L, ed. Pollination biology. Orlando, USA Academic Press, 151-161. Muller J. 1979. Form and function in angiosperm pollen. Annals ofthc Missouri Botanic Garden 66: 593632. P y e l MD,Harvey PH. 1988. Recent developments in the analysis of comparative data. Quarter& Review of Biology 63: 4 13-440. Pandey KK. 1971. Pollen size and incompatibility in Nicotiana. In: Heslop-Harrison J, ed. Pollen development and physiology. London: Butterworths, 31 7-322. Prcaton RE, 1986. Pollen-ovule ratios in the Cruciferae. Amcrican Journal of Botany 73: 1732-1 740. P h c k RB. 1987. Relationship among flowers, fruits and seeds. Annual Review of Ecology and Systematics 18: 409-430. Proctor M,Yco P. 1973. The pollination ofyowen. London: Collins. Qpella DC. 1984. Pollen-ovule ratios and hermaphrodite sexual allocation strategies. Evolution 38: 1148-1 15I . Ross-C~mig8. 1-1963. Drawings of British plants. Parts XV-XVIII.London: Bell. Salisbury EJ. 1942. The reproductive capan'& of plants; studies in quantitative biology. London: Bell. 248 W. D. J. KIRK ShunLer RU, cllle.hri.h KN, 1984. Does pollination efficiency shape the pollen grain to ovule ratio? Current Science 53: 751-753. Silvertom, J. 1989. The paradox of seed size and adaptation. Trends in Ecology and Evolution Ir 24-26. S m i t h CC, Fretwell SD. 1974. The optimal balance between size and number of offspring. American Naturalist 108: 499-506. Stanton ML. 1984. Seed variation in wild radish: correlation between propagule size and seedling performance. Ecology 65: 1105-1 112. Taylor TN,Levin DA. 1975. Pollen morphology of Polemoniaceae in relation to systematics and pollination systems: scanning electron microscopy. Grana 15: 91-1 12. Thompmm I(. 1990. Genome size, seed size and germination temperature in herbaceous angiosperms. Euolutionaty 7 ~ e n d sin Plank 4: 11 3-1 16. Whitehead DR. 1969. Wind pollination in the angiosperms: Evolutionary and environmental considerations. Evolution 23: 20-35. William. Ec, Rouse JL. 1990. Relationships of pollen size, pistil length and pollen-tube growth-rates in Rhododendron and their influence on hybridization. Sexual Plant Reproduction 3: 7-1 7. Willmon, MF,Burley N. 1983. Mate choice in plants: tactics, mechanisms and consequences. Princeton: Princeton University Press.
© Copyright 2025 Paperzz