Hereditas 86: 251-266 (1977) Effects of the Pallas in barley: phene analysis, overdominance, variability

AKE GUSTAFSSON, GUNNAR EKMAN AND INGEGERD DORMLING Institute of , University of Lund Semka AB, Sundbyberg The Phytotron, Swedish University of Agriculture, Stockholm, Sweden

GUSTAFSSON,A., EKMAN,G. and DORMLING,I. 1977. Effects of the Pallas gene in barley: phene analysis, overdominance, variability. - Hereditas 86: 251 -266 Lund, Sweden. JSSN 001 8-0661. Received May 1, 1977 Pallas barley is a high-yielding mutant induced by X-rays in Bonus barley and was among the first cereal mutants released into practice. It is characterized by a high lodging resistance and surpassed its parent variety Bonus, under good soil conditions, in yielding capacity. It has been widely cultivated in several European countries. It wasconsidered worth while to analyse the phene reactions of the mutant, its parent strain Bonus and their F,-hybrid (Pallas x Bonus) in the Stockholm phytotron under varied climatic conditions. Three photoperiods and two thermoperiods were chosen for study. Fifteen phenotypic traits (phenes) were analysed concerning generative and vegetative production, occur- rence and expression of dominance and overdominance, as well as phene variability (flexibility). Overdominance was common and often highly significant. Variability of individual phenes changed widely. In generative traits Pallas was more variable than Bonus and the F,-hybrid. Under certain climatic conditions, the hybrid, in its turn, was definitely less variable than the parents. An evident interaction of and climatic condition on productivity and variability was found. Ake Gustafsson, Institute of Genetics, S-223 62 Lund, Sweden

The barley mutant erectoides 32 (ert-k3’)was isolated In the year 1974, cross materials of Pallas x Pallas, in 1947 from Svalov’s Bonus barley. It became the Pallas x Bonus, F,, and Bonus x Bonus were pro- first wide-spread mutant variety in cereal species after duced by ordinary hybridization methods in the field. its official approval in 1958. Its high yielding potential In the following, the three materials are registered was established by Swedish, Danish, British and Irish respectively as PP, PB and BB. Pallas was chosen as results of cultivation (literature and discussions by the female hybrid parent in order to exclude a conta- GUSTAFSSON1963a and b, as well as by BORC et al. mination with selfed offspring in the F, materials. 1958, BORG 1959). A strict relationship between its Pallas is morphologically deviating from Bonus by its yielding ability and lodging resistance was shown to short and dense erectoides spikes (Fig. 1). The occur in official Danish trials (GUSTAFSSONand EK- mutant factor is most likely on the gene level. No MAN 1967). In the last decade, the mutant has gained sterility of genetic nature is found in hybridization. a new market in Spain, where it, as also its hybrid The mutant behaves in ordinary field experimentation descendant Hellas barley from Svalov, has covered as a recessive and segregates in a regular fashion. wide areas. Its parent variety Bonus, at a time both wide- spread and high-yielding, was originally produced by NILSSON-EHLE(see FROIER1954). The two , Material and methods mutant and parent, together with their F,-mono- hybrid offer suitable materials for studies on growth Three genotypes were tested in parallel phytotron and productivity in the Stockholm phytotron under experimentation: varied climatic conditions. Also features of domi- 1. Pallas x Pallas (PP) nance, overdominance and phenotypic variability 2. Pallas x Bonus (PB) could be included in the analysis of the data obtained. 3. Bonus x Bonus (BB) 252 A. GUSTAFSSON ET AL. Herediras 86 (1977)

Fig. 1. To the left: four spikes of Bonus barley, to the right: four spikes of Pallas barley (erectoides 32).

Among previously tried phytotron conditions, cf. GUSTAFSSONet al. 1975). Each genotype was three photoperiods were selected, known to give represented by maximally 10 individuals and distri- reproducible and rather normal growth effects (GUS- buted over the two trucks counted as separate units TAFSSON et al. 1974, 1975; DORMLINCet al. 1979, viz. in randomization. No root competition occurred. 1. 24 hours of light; with ca 22 000 lux at seedling level Border effects of plant position were not found.

2. 20 7, 9, ,, Fifteen variables (traits, phenes) were registered: 3. 16 ,, ,, ,, Two temperature conditions were applied. In Generative traits previous experiments they gave a rather low pheno- 1. Spike fertility, % typic variation with regard to seed setting and spike 2. 1000-kernel weight, g fertility 3. Kernel weight/plant, g 1. 20-10°C (20" for 16 hours, 10" for 8 hours) 4. Dry weight of spikes/plant (air-dried), g 2. 15-10°C (15" for 16 hours, 10" for 8 hours) 5. Number of kernels/plant The experiments were arranged in the form of a semi-randomized distribution of the three geno- Semi-generative traits types. Two trucks were used for a genotype compari- 6. Number of spikelets of first spike son, covering 2 x 14 pots, each truck being 0.25 mz in 7. Length of first spike, mm area. The plants were numbered from 1-28 (Fig. 2, 8. Number of heading tillers/plant Herediras 86 (1973 EFFECTS OF THE PALLAS GENE IN BARLEY 253

Fig. 2. 28 semi-randomized positions of three genotypes (PP, PB and BB) in six different photoperiod- thermoperiod arrangements.

Vegetative traits The traits measured are of the quantitative type. 9. Length of first culm to flag-leaf, mm Here, it may be pointed out that the pleiotropic 10. Length of first culm to spike base, mm differences, caused by the homozygous Pallas gene, 11. Number of shoots/plant are rather striking in field cultivation. In spite of the 12. Dry weight of straw (vegetative matter) quantitative changes involved, the mutant is of the per plant, g “macro”, not of the “micro”, type, as further indi- cated below. Traits of earliness The analysis of variance gives high significances for 13. Time from sowing to flag-leaf formation, days numerous individual traits: 14. Time from sowing to heading, days (1) with regard to genotype response in traits 2, (3), (15.Time to maturity, days) 4,(6),7,9,(10),(13),(14),i.e.especiallyin 1000-kernel weight, dry weight of spikes, spike length, culm length Thus, the materials analysed cover: 3 genotypes, to flag-leaf, 3 photoperiods, 2 thermoperiods, 15 variables at the (2) with regard tophotoperiodresponse in traits 1,2,6, end stage, with 56 PP, 58 PB and 54 BB individuals, 7, 9, 12, 13, 14, i.e. especially in spike fertility, 1000- which means 2520 measurements in all. Additional kernel weight, number of spikelets, spike length, culm measurements were continually carried out every 14th length to flag-leaf, dry weight of straw, earliness, day during the growth period. (3) with regard to thermoperiod response in traits 1,6, (7), 9, 13, (14), i.e. especially in spike fertility, number of spikelets, culm length to flag-leaf, earliness (time of 1. General processing of data: genotype, photoperiod, ’ flag-leaf formation), thermoperiod effects (4) with regard to interaction of genotype and photo- Trait 15, plant maturity, was excluded from the dis- periodin traits 2 and 11, i.e. in 1000-kernel weight and cussion owing to the difficult evaluation of the indi- number of shoots, vidual mixed genotypes. Traits 13 and 14 (time to (5) with regard to interaction of genotype and thermo- flag-leaf formation and heading) can be analysed with period in traits 2, 11 and 12, i.e. 1000-kernel weight, full accuracy, however. The number of measurements number of shoots and dry weight of straw, and used amounts to 2352. (6) with regard to interaction of photo- and thermo- Each plant and trait is considered one observation. period in all traits except for traits 7 and 9 (spike The data as processed in Table 1 suffice for a detailed length and culm length to flag-leaf). These two traits phenogenetic analysis. A noteworthy result is the high involve the two “macro” differences between Pallas degree of accuracy indicated for most conditions and and Bonus. traits, in spite of the low number of individuals of The high accuracy of phytotron cultivation, also in each genotype and condition (9-10 individuals). quantitative traits with small differences, is evidenced 254 A. GUSTAFSSON ET AL. Hereditas 86 (1977)

Table 1. Processing of genotype, photo- and thermoperiod influences on 14 variables in phytotron cultivation of mutant barley (Bonus x Bonus, Pallas x Bonus, Pallas x Pallas)

Variable I 2 3

Geno- Photo- Inter- Geno- Thermo- Inter- Photo- Thermo- Inter- type period action type period action period period action

1. Fertility - *** - - I** - *I* *** *** 2. 1000-grain weight *** *** ** *** - *** *** I** *** 3. Kernel weight * * - *** *** - - ~- *** 4. Spike weight ** - - *** *** - - ._ *** 5. No. of kernels - *** - - *** __ - ~- I**

6. No. of spikelets *** *** - * *** - *** I** *** 7. Spike length *** *** - *** *** - *** * __ 8. No. of heading tillers - *** - - __ __ - ~~ ***

9. Culm length flagleaf ** *** - ** ** - *** *** - 10. Culm length spike base * * - *** *** - - ~- *** __ - - .- 11. No. of shoots - *** *** *** *** 12. Dry weight straw - *** - - - *** I** I** ***

13. Flag-leaf formation *** *** - * ** - *** *** *** 14. Time of heading *** I** __ - * __ *** *** *I*

by this set of experiments. Of 126 individual differen- 3. Generative traits (1-5 in Table 1) ces no less than 59 lie below P-values of 0.001 and 14 -spikefertility (Table2). -As evidenced by more lie below 0.05, with either** or*. Table 1 both Dhoto- and thermoDeriod exert a -areat influence on the fertility in the conditions applied. All three genotypes are less fertile at the high temperature (20-10°C) than at the low one (15- 2. Detail analysis of individual traits loo). Pallas values at 15-10" are lower than those of A complete analysis of the 14 traits is not included in Bonus. In all six conditions, Pallas and Bonus possess the following, but each group of phenetic characters a lower fertility than their hybrid, with instances of will be represented. Bonus is throughout used as significance for PB:BB, and high significances for standard. PB:PP. The decrease of fertility is pronounced at 20-10" and 16 hours light. Trait 2 - 1000-kernel weight (Table 3). - This implies a character highly influenced by exterior Table 2. Spike fertility, %, at various thermo- and photoperiods of Pallas (PP), monohybrid (PB) and Bonus (BB) genotypes Table 3. 1000-kernel weight, g

Thermoperiod: 20-10" 15-10" Thermoperiod: 20-10" 15-10"

Genotype: PP PB BB PP PB BB Genotype: PP PB BB PP PB BB

Photoperiod Photoperiod 24 hours 75.7' 83.4" 81.5 88.9* 92.2" 91.1 24 hours 36.9*** 50.6" 50.5 49.5" 51.7" 50.8 20 hours 64.8" 74.0" 61.9 89.8" 92.7" 92.4 20 hours 36.3*** 46.9*** 41.7 44.0" 45.6" 45.2 16 hours 31.3" 31.9'*' 22.7 84.5*** 92.5" 91.7 16 hours 37.2*** 40.6" 42.8 39.6' 41.5*** 36.7 Herediras 86 (1977) EFFECTS OF THE PALLAS GENE IN BARLEY 255 conditions (AKERBERGand HAIDER1976). In the Table 4. Kernel weightlplant, g present case, striking genotype differences are in- volved. At the high temperature (20-10") Pallas is Thermoperiod: 20- 10" 15-10" inferior at every photoperiod to Bonus and the mono- hybrid. At the low temperature (15-10°) it is Genotype: PP PB BB PP PB BB approximately equal to Bonus at 24 and 20 hours of light and somewhat superior (*) at 16 hours. The Photoperiod 24 hours 7.64*** 10.23" 10.88 13.06' 14.76" 14.81 hybrid PB is superior to Pallas in all six conditions, 20 hours 9.65" 12.02" 11.77 13.97" 14.76" 15.34 to Bonus in five conditions, two of which are highly 16 hours 5.49" 7.98** 6.14 15.21" 16.46" 15.94 significant. The detail analysis agrees with the general behav- iour as evidenced in Table 1, where genotype, photo- Table 5. Number of spikelets of first spike period, thermoperiod and interaction effects are con- spicuous. Thermoperiod: 20-10" 15-10" Trait 3 - kernel weight per plant (Table 4). - In the analysis by GUSTAFSSONand EICMAN(1967) Pallas Genotype: PP PB BB PP PB BB was shown to be superior to 19 of 20 varieties tested, inferior to none, in official Danish yield trials. Bonus Photoperiod was included in the tests. However, it could also be 24 hours 27.6" 29.7*** 28.3 30.0" 29.4" 29.0 shown that the superiority of Pallas over Bonus was 20 hours 30.0" 31.6'' 30.4 33.3" 32.9" 31.8 connected with an increase in lodging resistance. In 16 hours 31.0** 34.1" 33.4 35.1* 36.6' 35.9 the phytotron experimentation reported here, every plant was grown in sterilized sand and all individuals were daily given low concentrations of Hoagland Trait 5. Number of kernels per plant. - The overall solution (100 mg nitrogen per liter water). Moreover, analysis of variance (Table 1) indicates no genotype the experiment was arranged in the form of a mixed influence but conspicuous photoperiod, thermo- genotype trial, without lodging and with no root com- period and interaction effects. petition. This does not correspond, of course, to the The six conditions result in kernel means of 264, behaviour in uniform genotype field cultures. In fact, 278 and 285 kernels per plant in PP, PB and BB Pallas reacts as a lodging-resistant "nitrogen ecotype" respectively, genotype differences without signifi- (GUSTAFSSON1954), in contrast to its easily lodging cance. The order of average productivity is parent. The overall analysis (Table 1) indicates a genotype 15-10", 16 hours: 406 kernels/plant and thermoperiod influence on kernel weight per ,, 20 ,, : 328 ,, plant. This is even better expressed in trait 4: air-dried ,, 24 ,, : 282 ,, spike weight per plant. 20--10",20 ,, : 269 ,, A closer study shows that Pallas is inferior to ,, 24 ,, : 208 ,, Bonus in every instance, in some cases significantly so. ,, 16 ,, : 163 ,, But the inferiority becomes less following a decrease in photoperiod. At 16 hours Pallas does not signifi- The three genotypes react in a similar way. The short cantly differ from Bonus. Apparently Pallas is less photoperiod of 16 hours and 15-10" gives the highest inferior at short than at long day. This fits in with its productivity. The condition of 16 hours and 20-10 is adaptability to South-European conditions. least productive. It is also obvious that the Pallas gene reacts in a similar way in the heterozygous state, where PB is significantly superior to Bonus at 16 hours and 4. Semi-generative traits (6-8 in Table 1) 20-10" and slightly superior at 16 hours and Trait 6 m number of spikelets offirst spike (Table 5).- 15-10", This result reflects the increase of hybrid In this trait there are definite genotype, photoperiod, fertility under every condition, but especially at thermoperiod and interaction effects. Interesting is 20-10" with 20 and 16 hours light. the pronounced increase of spikelet number in the The Pallas heterozygote is the highest-yielding monohybrid as compared to Bonus: it is superior in genotype in the series. The yield of 16.5 g/plant all six comparisons, in four of them significantly. (15-10°, 16 hours) corresponds to circa 9250 kilo- In four comparisons the monohybrid is superior grams per hectare. also to the mutant, presenting cases of clear over- 256 A. GUSTAFSSON ET AL. Hereditas 86 (1977)

0 0 0 0 mrn 0 0 100 0

0 0 0 0 0 0 0 0 0 0 0

0 50 0

Fig. 3. Spike lengths of Bonus and Pallas barleys at different temperaturecombinations and conditions under a photoperiod of 24 hours (phytotron cultivation 1967, cf. Fig. 4).

dominance at 20-10". Pallas is inferior to Bonus in Table 6 illustrates the striking differences in all four of six cases. comparisons between Pallas and Bonus. In both The number of spikelets is higher at 15-10' than thermoperiods, spike length increases when the at 20-10". It also increases with a shortening of the photoperiod is decreasing. The spikes of the mono- photoperiod. hybrid are in five cases shorter than those of Bonus Trait 7- spike length (Table 6). - In the introduc- ,(equal in one case). In all instances the monohybrid tion of this paper, it was pointed out that spike length has longer spikes than the mutant homozygote. is generally decreased in Pallas barley. In the present The degree of dominance of the normal allele varies comparison, total spike length was measured. Table 1 from circa 40 per cent (24 hours, 15-10°) to 100 per indicates a conspicuous genotype, photoperiod and cent (20 hours, 20-10"). The average value for all six thermoperiod influence but no interaction effects. comparisons amounts to 75 per cent. In Figs. 3 and 4, there are comparisons of spike length illustrating differences in a previous analysis of Table 6. Spike length, mm (first spike) Bonus and Pallas and using a wide range of thermo- periods. The photoperiods were 24 and 16 hours. Over Thermoperiod: 20-10" 15-10" the entire range of temperatures of varied duration, Genotype: PP PB BB PP PB BB there was a regular inferiority of the Pallas genotype. ~~ ~~ The heterozygote was not included in that ana- Photoperiod lysis. 24 hours 89.6*** 104.8" 107.0 91.4*** 97.0'. 104.7 Trait 8 -number of heading tillersperplant. -This 20 hours 93.4*** 107.0" 107.0 92.8*** 101.6*** 107.3 is a trait agriculturally more important than the pre- 16 hours 99.8*** 113.5" 117.2 96.4*** 109.3" 111.8 vious two. In contrast to them, this one possesses a Hrredilus 86 (IY77) EFFECTS OF THE PALLAS GENE IN BARLEY 257

0 mm #U 0 0 0 0 0 0 0 0 0 0

50

Fig. 4. Spike lengths of Bonus and Pallas barleys at different temperature combinations and conditions undera photoperiod of 16 hours (high temperature for 16 hours, low temperature for 8 hours).

pronounced capacity of phenotypic flexibility, often DORMLINGet al. (1975), it could be concluded, how- hiding existing genotypic differences. In this experi- ever, that “false” flag-leaves are often formed and mental series, as seen in Table 1, there is no genotype easily mistaken for “ordinary” ones under adverse difference noticeable by the analysis of variance. short-day conditions. Both traits ought therefore to be There occur, however, a distinct photoperiod effect generally considered. and a conspicuous photo-thermoperiod interaction. With regard to trait 9 (“flag-leaf’) there are evident The detail analysis does not present much more genotype, photoperiod and thermoperiod influences. evidence. There is a distinct photoperiod effect in all No interactions were found. In trait 10 (“spike base”), three genotypes and in both thermoperiods. Interest- the genotype-thermoperiod influence is more ing are four cases of heterozygote “underrecessive- pronounced, also with a clear photo-thermoperiod ness”, none of which, however, are significant. In one interaction. In both traits the genotype effect is simi- case the heterozygote is intermediate between the two lar. homozygotes, in one case insignificantly superior. In trait 9, the monohybrid is superior to Bonus in all six comparisons and to both Bonus and Pallas in five cases, two of which are significant (Table 9). 5. Vegetative traits (9-12 in Table 1) In trait 10, the heterozygote is superior to Bonus in Traits 9 and 10 - culm length to Jug-leaf and spike all six comparisons too, but with much higher base, mm. - In previous analyses, the flag-leaf significances, and to Pallas in five cases. In the condi- character was preferred to the trait of culm length to tion 20 hours and 15-10’, the Pallas homozygote is spike base. The process of heading was considered the superior genotype in both traits. especially fluctuating and highly influenced by The results relating to photo- and thermoperiods exterior conditions. After the study published by are confirmed by the detail analysis. The thermo- 258 A. GUSTAFSSON ET AL. Hereditas 86 (1977)

Table 7. Time from sowing to heading, days (means of Pallas is in all comparisons later than Bonus. The three spikes per plant) heterozygote is intermediate. The degree of domi- nance varies between 37 and 58%, with an average of Thermoperiod: 20-10" 15-10" 50%. It is interesting to note that with 16 hours day-light Genotype: PP PB BB PP PB BB Pallas seems to develop a little earlier at 15- 10" than at 20-10". This reverse behaviour has only a slight Photoperiod significance (*). Bonus and the hybrid are invariably 24 hours 47.5*** 45.8** 44.4 51.0" 50.4" 50.1 20 hours 55.7"" 54.0* 52.7 59.2*** 58.0" 56.7 later at the low temperature than at the high. 16 hours 69.9*** 67.1' 65.6 66.4" 67.4' 66.3 Trait 14 - time to heading (Table 7) shows approximately the same picture as the preceding one. The thermoperiod influence appears a little weaker, however. The previously mentioned reverse behav- Table 8. Generative efficiency iour of Pallas at 16 hours time is, in this trait, highly significant (***). Therrnoperiod: 20-10" 15-10"

Genotype: PP PB BB PP PB BB 7. Generative efficiency Photoperiod This term implies the quotient between generative and 24 hours 1.29 1.69 1.58 1.44 1.59 1 .60 20 hours 1.26 1.55 1.43 1.33 1.51 1 .57 vegetative production, i.e. grain yield versus yield of 16 hours 0.56 0.72 0.56 1.20 1.45 1 .32 dry matter of straw and leaves. A calculation has been made with regard to the three genotypes tested. As previously mentioned (p. 255), Pallas is less well period with a low day temperature gives rise to much adapted than Bonus to the prevailing phytotron conditions. Table 8 shows the results of calculation. longer culms than with high day temperatures. Interesting is the wider range of culm length under The obvious depression of the quotient at 20-10" the varied conditions dealing with the trait of length and 16 hours light is an expression of a strong vegeta- to spike base, viz. 673 to 1012 mm, than with the trait tive development and a low kernel yield under this of flag-leaf formation, viz. 750 to 970 mm, confirming condition. In all three photoperiods of this tempera- the great flexibility of the heading process in trait 10. ture, the hybrid shows the highest efficiency followed by Bonus and, then, by Pallas. A similar trend is also Trait 12 - dry weight of straw, g. - This trait found at 15-10" and 16 hours where the generative behaves approximately as trait 11 (total number of yield is quite high (cf. Table 4). shoots), but gives some clearer evidence of photo- period, thermoperiod and interaction effects. No genotype participation is found in either case (Table I). A few cases of significance occur under the higher 8. Dominance relationships temperature condition, where the mutant homo- The authors do not intend to review the field of zygote appears to be inferior both to Bonus and the dominance and overdominance. For previous studies monohybrid. In the lower temperature range, on the of barley mutants we refer to papers published contrary, the mutant homozygote gives some ten- by GUSTAFSSONet al. (1950), GUSTAFSSONet al. dency of superiority, at 20 and 16 hours of light, where (1975). The paper first mentioned stressed the need of the heterozygote seems to be "underrecessive". The comparing mutants, spontaneous or induced, in individual values are much higher at 15-10" than at a series of environments before definite judgments 20- lo", indicating the more vigorous vegetative could be reached with regard to their mode of produc- development at a low temperature. tivity and fitness. The terms "positive and negative heterosis" (STERN1948) are here replaced by the ex- pressions overdominance (superdominance, FISHER 6. Traits of earliness (13-14 in Table 1) 1918, 1932) and underrecessiveness (subrecessiveness; Trait 13 - time from sowing to flag-leaf formation. - equal to hybrid weakness, JONES1952; cf Fig. 5). This trait shows a considerable genotype influence When facilities became available for an appropriate (Table I), in addition to photo- and thermoperiod control of diverse climatic conditions (light, tempera- effects; there is also a strong interaction of photo- ture, humidity, nutrients) in clima chambers and and thermoperiods. Genotype interactions are absent. phytotrons, studies of gene effects could be analysed Hereditas 86 (1977) EFFECTS OF THE PALLAS GENE IN BARLEY 259

7i3e mtuntgune I over- ubmhune /hcoWk* rwe.c&e under- /eud/h,gr’o : bhonce d0mY;lcare ness neeessivwess Fig. 5. Different reactions of monohybrid PB compared to the original line BB and its recessive mutant PP: from over- or superdominance, via complete and incomplete dominance to complete recessiveness and under- or subrecessiveness.

in a more adequate way than previously possible. traits 2-5 (33 per cent) with Bonus dominance Also gene effects on individual plant characters (traits ranging between 24 and 99%. Underrecessiveness or phenes) could be studied in considerable detail. occurred in two cases (7 per cent). A general survey of a series of experiments carried out Semigenerative traits (6-8). - These traits show in the Stockholm phytotron will be published in the striking contrasts in expression. Variable 6 (spikelet near future, indicating the advantages but also weak- number) presents 4 cases of overdominance (three of nesses of such establishments. which are significant) and 2 cases of intermediate The three related genotypes discussed in this paper behaviour. Variable 7 (length of spike) is intermediate offer a possibility of further analysis of the dominance between the parents (with 42-almost 100% Bonus relationships of 14 quantitative traits in a strict dominance). Variable 8 (number of heading tillers) monohybrid of two pure lines (Bonus and its homo- has a third mode of reaction with four cases of under- zygous mutant Pallas). The data are presented in recessiveness, one case with intermediate behaviour Table 9. and one with weak overdominance. A few explanations of the table may be appropriate. Vegetative traits (9-12). - Also in this group of “Overdominance” (positive heterosis) simply means traits there are differences in behaviour. With regard that the hybrid is superior to both parents in one or to culm length (to flag-leaf, variable 9) five cases of several traits. In the case of “intermediate behaviour”, overdominance were found, two of which are signi- the monohybrid lies somewhere between the parents, ficant. The overdominance is even better expressed in from 0-100 per cent, calculated on the background variable 10 (culm length to spike base), with four of of the Bonus-Pallas differences. Underrecessiveness five overdominant cases highly significant. implies that the monohybrid is inferior to both In variable 11 (total number of shoots) there is one parents. In general Bonus is superior to Pallas; most weak case of overdominance, 2 cases of intermediate so in generative properties, least in vegetative ones. behaviour and 3 cases of underrecessiveness. Generative traits (1-5). - There is a definitely In variable 12 (dry weight of straw), two cases of skew distribution of the individual traits to the over- weak overdominance were found, three cases of dominant side (60 per cent of the cases), especially in intermediate behaviour and one underrecessive case the highly reactive trait of spike fertility (1) and in with some significance. 1000-kernel weight (2). Overdominance is least evi- Traits of earliness (13-14). - In 11 out of dent in variable 5 (kernel number). Five cases of high twelve cases there is an intermediate behaviour of significance were found. almost exactly 50 per cent Bonus dominance: in trait Intermediate behaviour was found in 10 cases of 13, between 37 and 57 per cent, with an average of 260 8. GUSTAFSSON ET AL. Hereditas 86 (1977)

Table 9. Analysis of dominance relationships in three barley genotypes (Bonus, Pallas and their F,-hybrid)

Variable’ Relative performance Dominance relationships

BBzPP PP>BB “Over- “Intermediate “Under- dominance” behaviour” reces- siveness”

1 4 2 6(1**)’ - 2 5 1 5 (I***) 1 3 6 - 3 (I**) 3 4 6 - 3 (I**) 3 5 5 1 1 (I**) 3 1-5 26 4 18 10

6 4 2 4 (I***, 2”) 2 I 6 - - 6 8 2 4 1 I 6-8 12 6 5 9

9 4 2 5 (I**, I*) 1 10 3 3 5 (2***, 2**) 1 II 4 2 1 2 12 4 2 3 9-12 15 9 7

13 63 - 6 14 63 - 5 13-14 12 - 11

’ Cf. Table I. ’ Within brackets cases of significance. ’ Bonus is earlier than Pallas.

49”/,, and in trait 14 (excluding a doubtful value), rence especially common at the thennoperiod 20- between 17 and 67 per cent, with an average of 52% lo”, with 11 instances of significance, compared with Bonus dominance. 3 cases of significance at 15-10’. The Bonus parent is earlier than Pallas in all 12 cases. Data of spike length (7) and earliness (13-14) differ considerably from those of other traits, espe- 9. Phenotypic traits and variability cially as regards fertility (l), 1000-grain weight (2), Individual quantitative traits show great differences number of spikelets (6) and culm lengths (9-10). in variability as seen in Table 10. The values essential- Conclusion. - In the analysed monohybrid geno- ly agree with those published in previous articles. type, overdominance was found in 18 of 30 generative In the table, coefficients of variation are given as cases, five of which are highly significant; in 5 of 18 percentages without decimals. In their paper, dealing semi-generative cases, three of which with high with “synthetic variables”, EKMANet al. (1975) also significance; and in 13 of 24 vegetative cases, six of considered the range of variation and the standard which are significant. deviation. The opposite behaviour, viz. “underrecessiveness”, was found in ten cases, one of which with a sign of weak significance. The traits showing overdominance A. Generative traits were generally displaced to the positive side, with no Spikefertility (trait 1) is the trait most easily reacting or few negative cases. The four traits with under- to changes in the environment. In all three genotypes, recessiveness, on the other hand, showed only one extremely low values appear at the thermoperiod of case of significant overdominance. 15-10°C. The genotype average for the three photo- A certain climatic differentiation with regard to periods is 2 per cent. At 20-lo”, the corresponding overdominance is suggested by the data, its occur- mean value amounts to 18 per cent, with variation Hereditas 86 (1977) EFFECTS OF THE PALLAS GENE IN BARLEY 261

Table 10. Variation coefficients (in %) of 14 phenotypic variables, 3 genotypes, 3 photoperiods and 2 thermo- periods For the sake of simplicity no decimals are included in the table, but the exact figures are used in the comparative data of the text

Photoperiod: 24 hours 20 hours 16 hours

Thermoperiod: 15---10 20- 10" 15-10" 20 ---I 0' 1 5 -I 0" 20-10'

Genotype: PP PB BB PP PB BB PP PB BB PP PB BB PP PB BB PP PB BB

Variable' I 3 I2 7 7 6 2 1113 715 4 2 2303046 1 3 6 4 7 4 5 6 3 4 II 5 5 6 4 9 8 8 6 3 13 13 10 27 9 II II 12 10 15 9 12 14 3 7 28 II 20 4 13 12 9 25 8 II II 12 10 13 9 12 14 4 7 20 8 15 5 14 17 13 23 8 14 15 14 12 15 12 7 16 6 9 30 12 22

6 434532831333222545 7 555533333524443537 8 22 18 13 23 13 13 20 11 14 21 15 10 17 12 15 18 10 7

9 587832355534327335 I0 6 5 5 11 3 3 2 4 4 'I 3 5 3 2 8 3 4 5 I1 20 18 II 21 10 II 16 8 19 16 13 18 16 12 12 21 16 14 12 8 9 7 14 9 7 8 II II 15 10 11 10 4 8 15 II 9

13 465556444544343534 14 455544444545333533

' Cf. Table I

coefficients rising when the photoperiods decrease. It is equally evident that Pallas is more variable than This contrast further exemplifies the "internal dis- Bonus (***) and the hybrid PB (***). The differences order" (endoataxy, DORMLINCet al. 1975, p. 268) in between Bonus and its Pallas-hybrid, however, are unsuitable conditions. contradictory and do not render any significance, if Also 1000-kernel weight (trait 2) gives rather low compared by a xz analysis (P>O.O5). The high coefficients of variation. Calculated, as in the pre- variation of the two homozygotes as compared with ceding trait, the coefficients are 5 per cent at 15-10" their hybrid is, however, suggestive at the high and 7 per cent at 20-10". temperature of 20-10" (PB< PP and BB in 11 With regard to kernel weight (trait 3), there are cases of 15) and especially at the photoperiod of higher coefficients of variation for all three genotypes. 16 hours (PB< PP and BB in 9 comparisons of 10, PP and BB, in partial contrast to PB, considerably P almost equal to 0.01**). increase their variability at the higher temperature. This change is most marked at 16 hours. B. Semi-generative traits A similar reaction is valid for spike weight (trait 4). The more or less adverse conditions of 20-10" show The two traits of spikelets of,.first spike and length of high values of variation in PP and BB, considerably ,first spike (6 and 7) give low coefficients of variation, smaller ones in PB. in most cases below 5 per cent. In contrast, the The trait of kernel number (trait 5) shows a lower number of heading tillers (trait 8) is more variable mean value at 15-10" and 16 hours compared with with coefficients between 7 and 23 per cent. 20--10", 16 hours. Hybrid PB is, on the whole, less With regard to the traits 6 and 7, the three geno- variable than PP and BB in the 20-10" environment types give approximately similar quotients of varia- and also at 16 hours 15-10. tion. In trait 8 (heading tillers), Pallas is definitely The striking increase in variability at the higher more variable than the other two genotypes, which temperature of 20- 10" as compared with the results are mutually rather similar. at 15-10' is confirmed by a X2-analysis (of the 45 The conditions of 20-lo", as compared with those comparisons). This gives a P-value less than 0.001***. of 15-10', even in combination with 16 hours, 262 A. GUSTAFSSON ET AL. Hereditus 86 ( 19777

m m do not change the extent of phenotypic variability in the semi-generative group of traits. m a C. Vegetative traits a a The traits relating to culm length (9 and 10) vary little under the applied conditions, without any distinct m m difference in variation coefficients relating to climatic environment or genotype. No evidence of any hetero- m "a zygous effects exists; nor is Pallas the more variable 9 homozygote. In trait 11 (number of shoots), the la coefficients of variation rise to a rather high level (up Wa to 21 per cent). In this trait Pallas is possibly, on an average, more variable than the other genotypes. In m m trait 12 (dry weight of straw), the variability is again decreasing with a slight indication of a higher Pallas m 9 a variability at 20-10".

maI -a D. Traits of earliness Both traits (13 and 14) possess low coefficients of variation, without any clear effects of genotype or environment, in spite of striking time differences in m P a flag-leaf formation and heading. la Wa E. Conclusion In the paper by DORMLINGet al. (1975, p. 267), it was shown that spike fertility is an easily reacting trait in changes both of photoperiod and thermoperiod. Other generative traits react less conspicuously. It was pointed out, in addition, that semi-generativeand vegetative traits are less influenced than the generative ones. m m This conclusion principally holds true in the present analysis also. The individual traits show considerable m mutual differences in variation. Changes in the envi- a 9 ronment especially influence the variability of genera- la tive traits and then, above all, of spike fertility. Semi- %a generative and vegetative, as well as traits of earliness, are more stable as regards degree of variability. How- ever, it ought to be emphasized that the three photo- periods used in this investigation are, on the whole, relatively optimal for growth and seed setting. No doubt, however, there occur genotypical effects on variability in the material. Under several condi- tions the Pallas genotype is more variable than the other two genotypes investigated. The hybrid of Pallas and Bonus often shows a remarkably small phenotypic variability. This is especially evident in the generative traits. At photoperiods shorter than 16 hours, for instance 12 and 8 hours (DORMLINGet al. 1979, the increase of phenotypic variability was con- siderable also in vegetative development. Even the Hrrediras 86 (1977) EFFECTS OF THE PALLAS GENE IN BARLEY 263

rather stable Bonus barley gave rise to generative and With regard to the mutual ratio of Bonus itself and vegetative anomalies at the extreme condition of 8 the monohybrid, it is obvious that the hybrid is the hours day-length. least variable genotype at 20- 10" of all photoperiods. With regard to gross morphological anomalies At 15-10" and 24 and 20 hours this difference dis- there occurred distinct genotype differences in reac- appears but is again conspicuous in the photoperiod tion. of 16 hours. These data, too, agree with the previously applied Xz-calculation. The semi-generative values 6-8 indicate no clear 10. Synthetic variables and degree of phenotypic genotype or environmental influence on the pheno- variability typic variability. All values lie on a rather low similar In a paper by EKMANet al. (1975), a method of multi- level, although on an average somewhat higher for variate analysis was worked out for the study of Pallas than for Bonus and the monohybrid. mutual genotype and environment influences on Concerning the vegetative variables 9-1 2, the data a series of quantitative traits. Three kinds of measu- suggest a consistently rather low variability of the ring the phenotypic variability were adopted: (I) the monohybrid. always below a synthetic value of 100. coefficient of variation, (2) the range of variation and The difference between Pallas and Bonus has largely (3) the standard deviation. In the comparisons re- disappeared. ferred to, the genotype influence was rather small Evidently there is a difference in reaction of pheno- (I.c., Table 3, p. 98). Thermo- and photoperiod, and typic variability between generative and vegetative their interactions, on the other hand, gave rise to high traits in the genotypes analysed in this experiment. values of significance for the eight traits selected and This is clearly seen both by examining the traits one by also for a synthetic variable, combining the values of one, and compositely, by using the synthetical all eight traits into one single numerical value of variable approach, modified as described above. It phenotypic variability. This was especially clear in should here be pointed out that this approach gives the case of range and standard deviation (I.c., Table 2, conclusive results in one direction only, in the follow- p. 96). ing sense. A linear function of the variables (coeffi- The synthetic variable approach was also applied in cients of variation) has indicated substantial differ- the material reported here, which is more homo- ences in the reaction of phenotypic variability. geneous, comprising only one parent variety and one A whole class of other linear or non-linear functions gene mutant with their simple monoheterozygote. might likewise indicate such differences, and any such The idea also in this case was to apply a composite function would then have served its purpose. Had formula for the evaluation of differences in variabili- the present linear function not indicated differences, ty, depending on genotype and photo- as well as this could not have been constructed to mean that no thermoperiod. Such an approach would involve the such differences exist, only that the present approach weighting of the coefficients of variation (Table 10) did not indicate them. The synthetic variable ap- together with the weights belonging to the synthetic proach, however, is an expression for a functional variable sum. However, this presupposes a uniform approach, which has, in a certain sense, the greatest "size" of the coefficients, which in itself is not possibility of indicating differences where they exist. possible to effect, since the coefficient of variation is without dimension. Consequently, a modified ap- proach is worked out. In Table 11, each of the variable groups 1-5,6-8 Results and discussion and 9-12 has been synthesized into a one-dimen- sional value weighting together the coefficients of Genotypes and traits. - The three genotypes Bonus variation using the products of the variable means and barley (BB), its released X-ray mutant Pallas (PP) and the synthetic variable coefficient as weights. their monohybrid (PB) have been cultivated in the As concerns the generative variables 1-5, it is Stockholm phytotron under six different, internally immediately evident that for all six conditions Pallas constant climatic conditions: three photoperiods and exhibits a higher variability than Bonus, as was also two thermoperiods. Fourteen (fifteen) quantitative shown by the direct xz analysis. The difference is less variables (traits, phenes) were measured, several well expressed at the otherwise highly influential through the entire vegetation period until ripening condition of 16 hours light and 20-10". In five of six time. Only the end results are recorded here. conditions, Pallas is more variable than the PB- The fourteen traits measured are divided into four hybrid; in most of the comparisons markedly so. categories: (I) generative traits (dealing with grain

17 264 b. GUSTAFSSON ET AL. Hereditas 86 (1977) properties and production), (2) semi-generative traits in this experimental series, there seems to be no doubt (indirectly affecting grain production), (3) vegetative that so-called hybrid weakness (JONES 1952) may traits (covering straw and leaf properties) and appear in fertile crosses also. (4) earliness (time to flag-leaf formation and heading). Spike length and earliness constitute deviating Analysis of variance. - The number of used traits with regard to dominance reaction. In all in- measurements amounted to 2352. The processing of stances (except one doubtful case) there is an inter- the data (Table 1) showed striking, often highly mediate action of the heterozygous gene, with a degree significantdifferences between genotype, photoperiod of dominance centering around 50 per cent. and thermoperiod effects and their respective inter- The distribution of values is oblique as to grain actions. Of 126 individual comparisons no less than characters (1 and 2 of Table 9), number of spikelets 59 gave P-values below 0.001*** and a sum of 73 (6) and culm lengths (9 and 10). P-values below 0.05*. A rough evaluation of phyto- Phene reactions and climatic conditions. - Refer- tron experiments as compared with field testings of ences are here made to Tables 2-8 and the corre- a similar nature indicates that the phytotron accuracy sponding text. is circa 10-20 times as high as that of good field Spike fertility is definitely decreased in all three experiments. genotypes, especially in Bonus at 16 hours light and Genotype differences. - Bonus is superior to Pallas 20- 10°C. Pallas is, concerning fertility, more sensi- in 26 of 30 generative, 12 of 18 semi-generative and tive to phytotron cultivation than Bonus, possibly in 15 of 24 vegetative traits, altogether in 53 of 72 trait connection with the nutrient factors (nitrogen bal- comparisons. In good soils, with a high level of ance). Thousand-kernel weight is highest at low tem- fertilizers, Pallas is less lodging than Bonus and also perature and long-day, and, in contrast, very low, higher yielding. It was previously denoted a “nitrogen especially in Pallas, at high temperature. The reac- ecotype” (GUSTAFSSON1954). In the phytotron tests, tions agree with those found by STOY(1976) and with no lodging occurring, Pallas is inferior to Bonus. AKERBERGand HAIDER(1976) for other materials. Its production is relatively seen higher at short-day High kernel yields are obtained at 15-10”, with the than at long-day conditions. This may partially highest value found in the monohybrid (16 hours, explain the high field productivity of Pallas in regions 15- 1Oo). The spikelet number is throughout increas- of cultivation more southern than devised for Bonus ing with a decrease in day length. Notice, for instance, itself, for instance, recently in Spain. the behaviour of Bonus (Table 5). A similar increase The hybrid PB is built on the one-gene (one-factor) is found also in spike length. difference of the two parents. Pallas itself, originally Generative efficiency. - This term involves the denoted as erectoides 32 (ert-k3’), is morphologically ratio between generative and vegetative production. rather a macro- than a micromutant, owing to its It is generally high under long-day conditions, with dense spikes and its more erect habit of growth. Its a superior value in the monohybrid at 24 hours light high specific productivity has been proved (GUSTAFS- and 20-10”. The very low values - below 1 - are SON 1963b). In spite of the very high productivity of recorded at short-day and high temperature (16 hours, the two parents, the monohybrid gives rise to definite 20-10”). Pallas is under phytotron conditions cases of overdominance (superdominance) in three generally less efficient than the other two genotypes. groups of traits (earliness excluded). In 36 of 72 Phenotypic variability. -- The trait of variability comparisons, the monohybrid is superior to both (phenotypic flexibility versus stability) is genotype parents, in 14 of them with rather high significances and environment influenced (DORMLINGet al. 1975, also over the better parent (Table 9). The overdomi- cf. MAYR1963, 1970). In the present analysis only the nance was well expressed in spike fertility (all six coefficient of variation, built on the standard devia- comparisons), 1000-kernel weight (5 of six compari- tion and the phene mean value, has been applied. In sons), number of spikelets (4 of six comparisons), previous studies the range of variation and the culm lengths to flag-leaf and spike base (10 of 12 standard deviation (variance) itself were also used in comparisons). The significances of overdominance comparison. are P<0.05 in 14 cases, PcO.01 in 13 cases and P It is evident that individual phenes show striking below 0,001 in 4 cases. differences in the degree of phenotypic variability In 72 comparisons, ten cases of “underrecessive- and that these differences change in size with the kind ness” (subrecessiveness)were found, in one case with of environment, as indicated by WENT (1953). Very slight significance (P<0.05). Its occurrence is espe- low figures of variability under optimal conditions are cially noteworthy in the number of heading tillers and shown by such traits as fertility, 1000-kernel weight, the number of shoots. Although not definitely proved spike and culm properties as well as earliness. In Hereditas 86 (1977) EFFECTS OF THE PALLAS GENE IN BARLEY 265 general, the generative group of traits markedly in- and Agriculture, as well as by the Knut and Alice Wallenberg crease their coefficients of variation at suboptimal or Foundation. actually unfavourable conditions. This is, for example, the case with fertility at 20-10”, especially with a day length of 16 hours. Such changes are less Literature cited pronounced in the semi-generative and vegetative traits and not noticeable at all in the case of earliness. AKERBERG,E. and HAIDER,T. 0. 1976. Climatic influences on However, there also exist clear genotype influences yield for summer cereals grown under northern climatic conditions. - Z. Acker- PJlanzenbau 143: 275-286 on variability. In the generative group of traits, Pallas BORG,G. 1959. Svalofs original Pallaskorn (Sv. 04032). nytt is less stable than Bonus and the monohybrid. The 2-radskorn, rontgenmutation ur Bonus. (Svalofs original last-mentioned genotype, with its one-gene hetero- Pallas barley (Sv. 04032), a new 2-row spring barley produced zygosity, shows a conspicuous decrease in generative by X-ray treatment of Svalofs Bonus Barley.) - Sveriges Utsadesforen. Tidskr., p. 72-96 variability (homeostasis, LERNER 1954). To some BORG,G., FROIER,K. and GUSTAFSSON,A. 1958. Pallas barley, extent this homeostatic balance of the heterozygote is a variety produced by ionizing radiation: its significance for found also in vegetative traits, especially at 20-10” plant breeding and . - 2 U.N. Int. Cot$ Peaceful with 24 and 20 hours light and at 15-10” and 16 Uses Atomic Energy Vol. 27 (P12468): 341-349 DORMLING,I., GUSTAFSSON,A. and EKMAN,G. 1975. Growth hours light. This becomes still more evident after the disorders and variability in phytotron-cultivated transformation of the individual coefficients by means barley. - Hereditas 79: 255-272 of synthetic variables (Table 11; cf. the text and EKMAN,G., GUSTAFSSON,A. and DORMLING,I. 1975. Pheno- EKMANet al. 1975). typic variability and the application of synthetic variables in barley. - Hereditas 81: 93-100 It is considered a general rule that phenotypic FISHER,R. A. 1918. The correlation between relatives on the variability is under a certain genotype control, but supposition of Mendelian inheritance. - Trans. Roy. SOC. genotype, climatic condition and individual phenes Edinburgh 52: 399-433 interact in specific ways, depending on the relative FISHER,R. A. 1932. The evolutionary modification of genetic phenomena. - Proc. 6 Int. Conf. Genet. I 165-172 strength of the different parameters involved. FR~IER,K. 1954. Aspects of the agricultural value of certain A series of phenetic changes have been found to barley X-ray produced and tested at the Swedish occur in the barley plant, from a complete balance of Seed Association, Svalof, and its branch stations. - Acta the growth characteristics (homeostasis, endotaxy), Agr. Scand. 4: 515-543 GUSTAFSSON,A. 1954. Mutations, viability and population via a set of internal and external disorders (mild structure. - Acta Agr. Scand. 4: 601-632 endoataxy) to a phene chaos with gross disorders and GUSTAFSSON,A. 1963a. Mutations and the concept of viability. abnormalities (severe endoataxy). Such disorders, - Recent Plant Breeding Research (Uppsala), p. 89- 104 ending in a phene chaos, were described in a previous GUSTAFSSON,A. 1963b. Productive mutations induced in barley by ionizing radiations and chemical mutagens. - publication by DORMLINGet al. (1975). The genotype Hereditas 50: 21 1-263 influence on degree of severity was obvious. GUSTAFSSON,A. 1972. The genetic architecture of phenotype In plants, an analysis of phenetic events is rather patterns in barley. - Induced Mutations and Plant Improve- easy to undertake, using controlled climatic condi- ment, IAEA, Vienna, p. 343-348 GUSTAFSSON,A. and EKMAN,G. 1967. Yield efficiency of the tions and isogenic seed or clone materials. In animals, X-ray mutant Svalof s Pallas barley. - Zuchter 37: 42-46 the study of phenetics has developed into a contro- GUSTAFSSON,A,, NYBOM,N. and VON WETTSTEIN,U. 1950. versial branch of science, dealing with species classi- Chlorophyll factors and heterosis in barley. - Hereditas 36: fication, evolution and phylogeny (cf. the discussion 383-392 GUSTAFSSON,A,, EKMAN,G. and DORMLING,I. 1974. Phyto- in MAYR 1965). Problems of phenetics are also tron ecology of mutant IV. Variability, photoperiod considered in the survey of evolutionary theory and . - Hereditas 76: 137--144 recently published by TIMOF!~EFF-RESSOVSKYand his GUSTAFSSON,A., DORMLING,I. and EKMAN,G. 1975. Phyto- co-workers (1975). tron ecology of mutant genes VI. Clima reactions of the eceriferum mutations cer-i16 and ~er~~.~ Hereditas80: 279- In this connection, the critical statement by 290 JOHANNSEN(1926, p. 165), concerning a dissolution JOHANNSEN,W. 1926. Elemente der exakten Erblichkeitslehre. of the phenotype into single separate units: phenes 3 ed. - Gustav Fischer, Jena (“sit venia verbo!”), ought again to be elicited. The JONES,D. F. 1964 (1952). Plasmagenes and chromogenes in heterosis. ~ In Heterosis, Hafner, New York, p. appearing phenotype is always the end result of a 224-235 chain of complex interactions within the living orga- LERNER,I. M. 1954. Genetic Homeostasis. - Oliver and Boyd, nism. Edinburgh MAYR,E. 1963. Animal Species and Evolution. - Harvard Univ. Press, Cambridge, Massachusetts

Acknowledgments. ~ Financial support of this project has MAYR,E. 1965. Numerical phenetics and taxonomic theory. - been rendered by the Swedish Research Council of Forestry Syst. Zool. 14: 73-97 266 A. GUSTAFSSON ET AL. Hereditas 86 (IY77)

MAYR,E. 1969. Principles of Systematic Zoology. - McGraw- different barley genotypes. -- Barley Generics I11 (Garching Hill, New York 1973, p. 641-648 MAYR,E. 1970. Population, Species and Evolution. - Hurvurd TIMOFEEFF-RFSOVSKY.N. V., VORONCOV,N. N.and JABLOKOV, Univ. Press, Cambridge, Massachusetts A. N. 1975. Kurzer Grundriss der Evolutionstheorie. - STERN,C. 1948. Negative heterosis and decreased effectiveness Gustuv Fischer, Jena of alleles in heterozygotes. - Genetics 33: 215-219 WENT,F. 1953. Gene action in relation to growth and develop- STOY,V. 1976. Source and sink properties as related to yield in ment. - Proc. Nar. Acad. Sci. 39: 839-848