Heredity 56 (1986) 25—31 The Genetical Society of Great Britain Received 8 March 1985

The inheritance of in Sinapis arvensis L.

K. Garbutt* and J. R. Witcombet School of Plant , University College of North Wales, Bangor, Gwynedd, U.K. Selection for dormant and non-dormant seed in Sinapis arvensis was carried to the seventh and fourteenth generation, respectively. Crosses between the dormant and non-dormant lines clearly showed both a maternal and an embryonic component of seed dormancy. A model for the number of alleles controlling dormancy was constructed and tested. The maternal component of dormancy was shown to be controlled by a single locus with two alleles, the dormant allele being dominant to the non-dormant. No clear picture of the control of the embryonic component of dormancy was found.

INTRODUCTION shown a clear response to selection for non- dormancy in several populations. They also Therehave been several reports of genetic control demonstrated a clear difference in germinability of factors affecting seed dormancy in plants. between threshed and unthreshed grains, indicat- Honing (1930) showed a genetic basis and a dis- ing the presence of inhibitors in the lemma and tinct maternal effect in the light dependence of palea, and hence a maternal component to control behaviour in tobacco. Harper and of dormancy. McNaughton (1960) found evidence of genetic There have been two recent attempts to estimate control of seed dormancy in species of Papaver. the number of loci involved in the control of germi- New (1958; 1959; 1961) has shown a strict genetic nation. Guenther and Borriss (1975) tried unsuc- control of seed coat types associated with differing cessfully to estimate the number of loci controlling germination rates in Spergula arvensis indicating seed dormancy in Antirrhinum spp. concluding that a strict maternal control, via the seed coat, over dormancy was controlled by a polygenic complex. seed dormancy. However, Jana, Acharya and Naylor (1979) have Naylor and his co-workers have shown genetic shown that at least three loci control after-ripening variability in seed dormancy in Avena fatua in in Avena fatua. response to several environmental variables In the field, Sinapis arvensis populations applied both whilst the seed is maturing and during usually show two "flushes" of germination each germination tests (Naylor and Jana, 1976; Naylor year, one in spring and the other in autumn. and Fedec, 1978; Sawhney and Naylor, 1979). Jam Edwards (1980) suggests that the autumnal popu- (1982) has also demonstrated genetic components lation results from the non-dormant fraction of the of seed dormancy for Arena barbata and Tr(folium seed produced in the spring. In the field, S. arvensis hirtum but in the same study was unable to demon- germination is controlled to some extent by strate a significant genetic component of dormancy environmental factors. Edwards (1980) found in Bromus mollis. germination to be associated with soil temperatures Hilu and de Wet (1980) have investigated the of 44°C at a depth of 10 cm and with high rainfall genetic basis of dormancy in Eleusine sp. and have in spring and simply with adequate moisture in autumn. There is also evidence that at least some Present addresses: *TheBiological Laboratories, Harvard Uni- seed of S. arvensis require light to germinate versity, 16 Divinity Avenue, Cambridge, MA 02138. (Frankland, 1976). In the laboratory, dormancy is t ICRISAT, Pantancheru P.O., Andhra Pradesh 502 324 India. broken by the application of gibberellic acid 26 K. GARBU1T AND J. R. WITCOMBE (Witcombe and Whittington, 1972; Edwards, the parent plants has a seed coat of generation P1 1976). and embryo of generation pe,andthe plants grown The presence of a germination inhibitor and its from this seed will be generation peSeedfrom a mode of action in of S. arvensis was reported cross between two plants of pe generation will in Edwards (1968c). This inhibitor is apparently have Sgenerationseed coat and F generation produced by the seed coat (testa) (Witcombe, embryo. Hiliman and Whittington, 1969) and seems to be All plants were grown in a greenhouse in 10 cm present in the seed coat of all seeds in apparently plastic pots filled with John Innes No. 1 compost. similar quantities (see Edwards, 1968a for a Supplementary light to give a daylength of 16 hours detailed description of seed anatomy). The inhibi- minimum was provided by mercury vapour lamps tion is broken down by oxidation (Edwards, 1969) and temperature was maintained at a minimum of and the effect of the seed coat in dormant seed 18°C. Treatments of fungicide and insecticide were appears to be to prevent this oxidation taking place applied as necessary. by being impermeable to oxygen (Edwards, All pollinations were made in the same manner 1968b). The seed coat is clearly of maternal origin whether within or between lines. Dehiscing anthers and hence control of dormancy is maternal. There were removed from the male parent and placed in is evidence, however, of a further level of dor- a watch glass; a flower bud about to open, but not mancy control, possibly unrelated to this maternal yet showing the stigmatic surface was opened, and effect, and located in the embryo (Edwards, 1968c; emasculated and the newly exposed stigmatic sur- Witcombe, 1971). This points to one of the major face loaded with pollen from the anthers of the problems in studies of the inheritance of seed male parent. The pedicel was marked with paint dormancy. The seed combines parts of three dis- and 24 hours later a further application of pollen tinct generations (and genotypes)—the seed coat was made. (maternal), the (filial but with one These crosses produced generationseed, paternal and a variable number of maternal 200 individuals of which were grown on and genomes) and the embryo (filial). In the ripe seed allowed to cross at random to produce F7F of Cruciferae, however, the endosperm has disap- generation and seed 200 individuals of which, peared leaving a seed composed of two generations when grown on and crossed, produced FF seed. of tissue, seed coat and embryo. A further set of crosses was made between pe The purpose of the following investigation was plants. These comprised 50 crosses between dor- to determine the manner in which dormancy was mant males and females, 50 crosses between non- controlled genetically in Sinapis arvensis. dormant males and females, and 50 crosses between dormant males and non-dormant females. This was to confirm that the differences observed MATERIALSAND METHODS between the germination rates of seed from non- dormant maternal parents from the p1pe and the Inorder to establish the genetic basis of seed PF generations were, in fact, real and not a dormancy, two selection lines, one for dormancy function of a random difference between genera- (D) and one for non-dormancy (N.D.), were estab- tions. lished from a natural population collected in At each generation, pspePSFC Nottinghamshire by J. R. Witcombe (Witcombe germination tests were carried out on the seed in and Whittington 1972). These lines were taken to closed petri dishes each containing one sheet of the SlO generation for the ND. line and the S5 9.0 cm filter paper. Initially each dish received 5 ml generation for the D line by J. R. Witcombe and of distilled water. Thereafter, all dishes received K. Daniels. The present authors continued distilled water as necessary. The germination tests selection using the same methods as Witcombe were carried out at 25°C, with 24 hour light. The and Whittington (1972) and Witcombe (1971) to lids of the petri dishes were lifted daily to prevent the S14 generation for the N.D. line and S7 the build-up of gases which might inhibit germina- generation for the D line. At this point, 200 tion. Each dish contained 10 seeds unless otherwise reciprocal pairs of crosses were made between the stated. lines and seed produced to the third filial At the conclusion of each test (14 days), the generation. seeds that remained ungerminated were treated As mentioned in the Introduction, the seed is with high levels of G.A. (up to 3000 ig/Iitre) to made up of two distinct generations. Thus a special stimulate dormant seeds into germination and notation will be used. The seed which produces thereby detect non-viable seed. Gibberellic acid INHERITANCE OF SEED DORMANCY 27 was applied in an aqueous solution at pH 7. All generation S7 and then tended to reach a plateau, G.A. solutions were made up freshly before an though with a continued slight increase in germina- experiment and were used immediately. tion and decline in standard error (calculated from With the exception of FF bulk germination arcsine transformation) between S7 and S14 tests were performed (no record of individual plant (fig. 1). performance being kept). The results of the bulk Table 1 gives the mean percentage germination tests were based upon as many seeds as possible and 95 per cent confidence limits for the two lines (between 1000 and 2000 seeds) with only a small at the last generation of selection and the control sample being kept in reserve. Confidence intervals population. The N.D. line is clearly different from were obtained using tables from Rohif and Sokal the others. The difference between the control and (1981), when more than 1000 seeds were used the the D line is small (OO5>p>O0l) from a t-test limits for a sample of 1000 were used as a conserva- on arcsine transformed data. tive estimate. Tests of FF seed were based on Neither 100 per cent germination in the absence individual plants, five petri dishes of 20 seeds being of G.A., nor 0 per cent germination could be set up for each of 100 FF plants. All these tests obtained from the selection lines. Heritability for were carried out under standard conditions as pre- dormancy had been found by Witcombe (1971) to viously described and in distilled water. be approximately 20 per cent, so this failure has been attributed to environmental factors such as the position of the micropyle during germination, RESULTSAND MODEL or damaged seed coat. The differences observed (table 2) in the germi- Selectionwas continued in the N.D. line to nation behaviour of seed derived from different generation S14 and in the D line to S7. The maternal parents in lines which are expected to response to selection in the D line is not as obvious have the same embryo genotype (F generation) as that in the N.D. line and initially a rise in indicates that some part of the seed other than the germination rate was observed (fig. 1); this may embryo has a profound effect upon germination be due to the release of concealed variation during behaviour. the first stages of selection. The non-dormant line The work of Edwards (1968b) and Witcombe, appeared to respond sharply to selection up to Hiliman and Whittington (1969) indicate that it is 90

8(

SO S2 S4 S6 S8 Sb S12 S14 Generation

Figure1 The mean percentage germination of 14 generations of selection for non-dormant seed (0),sevengenerations for dormant seed (S),andof the unselected control population (a).Barsindicate two Standard Errors. 28 K. GARBUTT AND J. R. WITCOMBE Table 1 Themean percentage germination and 95 per cent differences between the seed derived from the confidence interval for the non-dormant (N.D.) and dor- mant (D) selection lines and the unselected control popula- N.D.xN.D. cross and seed from the Female tion of Sinapis arvensis N.D.xMale D cross (signficant at p

Table 2 The seed generations with their maternal phenotype Model (where known) and mean germination percentage (with 95 From per cent confidence interval) resulting from the ND. x D the results in this section, we know that the cross seed coat, and hence the maternal genotype, has an effect upon seed dormancy. It has also been Seed Maternal Mean % demonstrated that the embryo genotype, at least generation phenotype germination when associated with a non-dormant seed coat, LU has an effect upon dormancy. In the light of these P1P Dormant 2 1—3 results a working model is suggested where two PSIP Non-dormant 84 82—86 loci, each with two allelic genes control dormancy. PSF Dormant 7 6—9 One locus controls that character of the seed coat P'F Non-dormant 58 55—62 — 15 13—17 which inhibits germination with alleles I (inhibit- FF ing germination) and allele i (not inhibiting germi- nation). I is considered dominant to i. A second locus affects embryo dormancy, the allele D caus- the seed coat rather than the endosperm which is ing embryonic dormancy and d not causing embry- the important factor. As the testa is derived from onic dormancy. No dominance effect is postulated maternal material the results obtained are at least, at this locus. Although it is probable that the con- in part, a function of the maternal genotype. trol of dormancy in the embryo is more complex The results obtained for PSF seed show that than this model would suggest, no clear predictions an embryo effect is present as well as a maternal could be made about the embryonic component effect (table 2). In the case of seed with an N.D. so we chose this model for reasons of notational maternal parent, there was a significant difference simplicity and parsimony. Given the two parental between the germination precentage of the parents lines were homozygous, we can now produce p1pe generation and the offspring PSF genotypes for each generation of crossing (table 4). generation. If the only factorsaffecting If we assume, as seems reasonable, that the germination were maternal, these germination effect of I is dominant and only slightly modified percentages should have been the same. by the effect of embryo genes (Dd) while ii is To confirm this result, a further set of crosses altered more by embryo genes, the overall percen- between and within dormant and non-dormant tage germination of FF seed will be an almost lines was set up (see Materials and Methods). The meaningless value. However, it is predicted that results of these crosses are given in table 3. The seed coat genotypes II, Ii and ii will be in the ratio of 1:2: 1, and if I is dominant it is expected the 75 per cent of the F generation plants will produce Table 3 The germination percentages and their 95 per cent phenotypically dormant seed and 25 per cent confidence intervals of seed resulting from within and phenotpyically non-dormant seed. between line crosses of the N.D. and D lines In the recording of results of the germination Female parent Male parent Seed mean % germination test on FF seed, all plants which produced seed which had a germination percentage over 40 per LU cent were deemed to be phenotypically non- Dormant Dormant 8 6—11 dormant and the seed coat assumed to have seed Non-dormant Non-dormant 79 75—82 Non-dormant Dormant 53 57 coat genotype ii, based on the results of the 49— germination tests. INHERITANCE OF SEED DORMANCY 29 Table 4 The seed genotypes thatitis suggested occur in the P'1P, and FFseedgeneration

Seed Maternal % generation phenotype Germination Genotypes

PSP Non-dormant 84 Seed coat ii Embryoi i dd Dormant 2 Seedcoat II Embryo II DD PF Non-dormant 58 Seed coat ii Embryo Ii Dd Dormant 7 Seedcoat II Embryo Ii Dd FF — 15

Ii Ii IiIiIiIiIiIiIi 111111IiIiIiiiiiI DDDd ddDDDdddDDDddd As a result of germination tests on the FF was used for convenience of notation rather than seed of 100 plants, 21 plants were found to be from any feeling that it is necessarily correct. phenotypically non-dormant compared with an Further investigations are necessary to identify the expectation of 25non-dormantplants (x= true nature of the genetic control of dormancy in 0853 N.S.). Thus the FF seed behaves as the embryo. What has emerged clearly from this predicted by the model. work is the overriding importance of the maternal Figure 2 shows the divergence of the dormant- genotype in seed dormancy and that at least two type and non-dormant-type plants in the F gene loci control dormancy in S. arvensis. generation. The range of variation shown by the There have been several mathematical and com- non-dormant plants may indicate that the embry- puter models of seed banks and their effects upon onic component is more complex than the model fitness (e.g., Cohen, 1966; Charlesworth, 1974; suggests. However, environmental factors can and Templeton and Levin, 1979). It seems to be agreed do affect germination rates and as, at present, it is that one of the most important functions of the not possible to distinguish between these effects seed bank is to reduce the fitness uncertainty pro- the model is retained for reasons of parsimony. duced by cyclical and random environments and thus prevent the "tracking" of temporary adaptive optima. However, dormancy itself may be affected DISCUSSION by selection (as has been demonstrated here) and is not necessarily protected from random environ- Thismodel is a working base from which to pro- mental variations in the way that dormancy pro- ceed. From the various tests and the rate at which tects other characters. The effect of a maternal the non-dormant line responded to selection, it influence in the dormancy phenotype of seed will seems clear that the action of the seed coat in be to dampen the tendency to "track" environ- inhibiting germination is under major gene control. mental change, in much the same way as the seed The expression of seed coat dormancy is modified, bank itself dampens the response of a population. particularly in the case of the recessive allele, by This may help to explain the success of S. arvensis the alleles present in the embryo. as an agricultural weed and its survival under a The number of the genes expressed by the number of cultivation regimes. Edwards (1980) embryo is not clear from this work and the single showed that plants which germinate in the autumn locus two allele system proposed, while plausible, are prevented from reproducing by autumn 30 K. GARBUTT AND J. R. WITCOMBE 50

4

30

20

Il0

l0 20 30 40 50 60 7080 90 Percentage Germination of Seed per Plant Figure 2 The range and frequency of germination percentages shown by the seed (FF) and F plants. Plants with germination percentage less than 40 percent were considered dormant genotypes; those above 40 percent were considered non-dormant genotypes.

ploughing; yet because of the large number of REFERENCES seeds in the seed bank, some at least will produce non-dormant seed in those years when autumnal CHARLESWORTH,B. 1974. Selection in populations with over- lapping generation. VI. Rates of change of gene frequency ploughing is late or does not occur, and there will and population growth rates. Theoretical Population Bio- be an autumnal "flush" which will reproduce. logy, 6, 108-133. Spring ploughing will have a similar effect upon COHEN, D. 1966. Optimising reproduction in a randomly vary- those individuals which germinate in the spring, ing environment. Journal of Theoretical Biology, 12, 119- 129. but will also tend to bring up to the surface buried EDWARDS, M. M. 1968a. Dormancy in seeds of Charlock seed which will germinate later. As it requires only (Sinapis arvensis). 1. Developmental anatomy of the seed. one good year in 11 to maintain the seed popula- Journal of Experimental Botany, 19, 575-582. tion (Edwards, 1980) and with the prevalence of EDWARDS, M. M. 1968b. Dormancy in seeds of Charlock mixed agricultural practice in those areas where (Sinapis arvensis). 2. The influence of the seed coat. Journal of Experimental Botany, 19, 583—600. S. arvensis occurs, it is unlikely that a population EDWARDS, M. M. 1968c. Dormancy in seed of Charlock (Sinapis would be removed by cultivation alone. arvensis). Occurrence and mode of action of an inhibitor Thus, in S. arvensis we find a genetic system associated with dormancy. Journal of Experimental Botany, that controls dormancy not only by imparting a 19, 601-610. range of dormancy types, but also of a nature (the EDWARDS, M. M. 1969. Dormancy in seeds of Charlock (Sinapis arvensis). 4. Interrelationships of growth, oxygen supply element of maternal control) that protects from and concentration of inhibitor. Journal of Experimental over-reaction to selection by temporary environ- Botany, 20, 876-894. mental perturbations, and appears to be a valuable EDwARDS, El. El. 1976. Dormancy in seeds of Charlock (Sinap- sis arvensis). Early effects of Gibberellic acid on the syn- mechanism in adaptation to agricultural systems. thesis of amino acids and proteins. Plant Physiology, 58, 626-630. EDWARDS, MIRIAM. 1980. Aspects of the population ecology of Charlock. Journal of Applied Ecology, 17, 151-172. Acknowledgments The authors would like to thank Professor FRANKLAND, B. 1976. Phytochrome control of seed germina- J. L. Harper, Professor F. A. Bazzaz, Dr P. J. Murphy and an tion in relation to the light environment. Proceedings of anonymous reviewer for their comments and advice, and C. the 22nd Easter School in Agricultural Science, University Garbutt for her help with the crosses. This work was supported of Nottingham, Smith, H. (ed.) Butterworths, London, pp. by a Science Research Council grant to K. Garbutt. 169— 188. INHERITANCE OF SEED DORMANCY 31

GUENTHER, E. AND BORRISS, H. 1975. The genetic control of NEW, J. K. 1958. A population study of Spergula arvensis. Part germination of Antirrhinum seeds.2.The genetic basis of 1. Two dines and their significance. Annals of Botany, 22, influence of Gibberellic acid on the germination of 457-477. Antirrhinum seeds. Biologisches Zentralblatt, 94, 45—61. NEW, J. K. 1959. A population study of Spergula arvensi. Part HARPER, J. L. AND McNAUGHTON, I. H. 1960. The inheritance 2. Genetics and breeding behavior. Annals of Botany, 23, of dormancy in inter- and intraspecific hybrids of Papaver. 23—33. Heredity, 15, 3 15—320. NEW, J. K. 1961. Biological fluora of the British Isles. No 76 HILU, K. W. AND DE WET, J. M. j.1980.Effect of artificial Spergula arvensis. Journal of Ecology, 49, 205-215. selection on grain dormancy in Eleusine (Gramineae). Sys- ROHLF, F. J. AND SOKAL, R. R. 1981. Satistical Tables. W. M. tematic Botany, 5, 54-60. Freeman and Co. Oxford, 219 pp. HOMING, J. A. 1930. Nucleus and Plasma in the heredity of the SAWHNEY, R. AND NAYLOR, J. M. 1979. Dormancy studies in need of light for germination in Nicotiana seeds. Genetica, seed of Avenafatua. 9. Demonstration of genetic variability 12, 441—468. affecting the response to temperature during seed develop- JAIN, S. K. 1982. Variation and adaptive role of seed dormancy ment. Canadian Journal of Botany, 57, 59-63. in some annual grassland species. Botanical Gazzette, 143, TEMPLETON, A. R. AND LEVIN, D. A. 1979. The evolutionary 101-106. consequences of seed pools. American Naturalist, 114,233- JANA, S., ARCHARYA, S. N. AND NAYLOR, J. M. 1979. Dormancy 249. studies in seed of Avena fatua. 10. On the inheritance of WITCOMBE, i. R. 1971. Genetic analysis of dormancy, germina- termination behaviour. Canadian Journal of Botany, 56, tion and growth in three Cruciferous species. Ph.D. Thesis, 1663—1667. University of Nottingham. NAYLOR, 1. M. AND FEDEC, P. 1978. Dormancy studies in seed WITCOMBE, J. It., HILLMAN, J. AND WHJYrINGTON, w. .1969. of Avena fatua. 8. Genetic diversity affecting response to Growth inhibitor in the seed coat of Charlock. Nature, temperature. Canadian Journal of Botany, 54, 2224-2229. 222, 1200. NAYLOR, J. M. AND lANA, S. 1976. Genetic adaptation for seed WITCOMBE, J. R. AND WHITIINGTON, w. .i. 1972. The effects dormancy in Avenafatua. Canadian Journal of Botany, 54 of selection for reduced dormancy in Charlock (Sinapis 306—3 12. arvensis). Heredity, 29, 37-49.