How was rice differentiated into indica and japonica?

The evolutionary dynamics of the indica-japonica differentiation was studied from
the viewpoint of population genetics. Indica and japonica are distinguished by
genes and characters associated with each other nonrandomly. In indica/japonica
hybrid progenies, the same direction of gene and character association found
among the cultivars was generally observed. This means a trend toward the
restriction of recombination among several independent loci. Accordingly, intermediate
types between indica and japonica are relatively infrequent, even if natural
hybridization occurs frequently between them. lndicas and japonicas are isolated
by the restriction of recombination in the hybrids.

Two subspecies or ecogeographic races of common rice, indica and japonica, are
represented by genes and characters associated nonrandomly among many cultivars
(gene and character association). Oka (1958) defined the indica and japonica types as
two varietal groups having associations of genes or phenotypes in contrasting states of
phenol reaction ( Ph/ph ), apiculus hair length, KClO 3 susceptibility, and tolerance for
cold and drought. Glaszmann (1987) reported that two major varietal groups represented
by associations of alleles at 15 isozyme loci largely corresponded to the indica
and japonica types defined by Oka (1958).
Yet the causal factor of the indica-japonica differentiation among rice cultivars
remains unknown. To elucidate the factors causing nonrandom association of genes
and characters that result in the indica-japonica differentiation among cultivars, the
pattern of their associations was studied in hybrid populations. Here I describe patterns
of association in 12 genes and characters among many cultivars, and also in F 2 and F 5
populations derived from an indica/japonica cross, and discuss the factors causing such
nonrandom associations.
Materials and methods
Two hundred cultivars and 4 single seed descent populations derived from an indica/
japonica cross were tested to examine 12 characters and genes.
45
Plant materials
A sample of 200 native cultivars collected from various localities in Asia was used to
represent varietal variation occurring in nature. The cultivars were classified into
indica and japonica by the method described here.
Hybrid populations used in this study were derived from the cross Acc. 419 (indica)/
Acc. 504 (japonica). Both parents were included in the varietal sample. Acc. 419 was
developed by pureline selection in India. Acc. 504 is the Taiwanese cultivar Taichung
65 (T65), from a cross between two Japanese native cultivars. Acc. 419 is a typical
indica, and T65 a typical japonica. They have different alleles at a number of loci, and
different phenotypes.
The F 2 population consisted of 200 individuals. The F 3 and F 4 populations were
raised by the single seed descent method in which seed for the next generation is
prepared as a bulk of a single seed from each plant of the previous generation. The F 3
and F 4 populations consisted of 188 and 172 plants, respectively. The F 5 population was
raised by bulking 2 seeds from each F 4 individual, and 300 plants were randomly
chosen for analysis.
Genes and characters examined
All F 2 and F 5 plants were examined on a single-plant basis for phenol reaction ( Ph/ph,
chromosome 4), susceptibility to KClO 3 at the two- or three-leaf stage (genes
unknown), and apiculus hair length (in millimeters) to classify them as indica or
japonica. The same characters were also recorded in the 200 cultivars. Measurement
methods for these characters are described by Sato et al (1986). For quantitative
numerical evaluation of the indica-japonica differentiation, a discriminant score (Z)
was calculated for each cultivar by combining three characters as follows:
Z=Ph + 1.313 K – 0.82 Hr – 1.251
where Ph, K, and Hr indicate phenol reaction, KClO 3 susceptibility, and apiculus hair
length, respectively. Ph is 1 if positive or 0 if negative. K varies from 0.0 (most
resistant) to 2.0 (most susceptible). Hr value is given in millimeters.
The hybrids and cultivars were also examined for pericarp color ( Rc/rc, chromosome
7), apiculus color ( C/c, chromosome 6), hull color (black or straw, complementary
action of Ph, Bh-a, and Bh-b ), and awn (gene[s] unknown), which segregated in
the present cross. Furthermore, they were examined for five enzyme-encoding loci that
also segregated in the hybrid populations: Est-2 (chromosome 6), Pgi-2 (chromosome
6), Amp-2 (chromosome 8), Cat-1 (chromosome 6, but independent of Est-2 and Pgi-
2 ), and Acp-1 (chromosome 12). Detailed descriptions of the methods of isozyme assay
are given in Ishikawa et al (1987).
Results
Character and gene associations in the varietal sample
Correlations among Ph, K, and Hr among the cultivars are indicated in Figure 1. K
values showed continuous variation, but seemed to be divided into susceptible (W) and
46 Y. -I. Sato
1. Relations among phenol reaction, KClO 3 susceptibility, and apiculus hair length. R = resistant, W =
susceptible to KClO 3 , L = long, S = short apiculus hair length, o = negative phenol reaction, = positive.
How was rice differentiated into indica and japonica? 47
resistant (R) types. Hr values showed continuous and unimodal variation. However,
this character is controlled by the Aph locus; Aph carriers mostly have apiculus hairs
longer than 0.7 mm, while aph carriers have ones shorter than 0.7 mm (Sato 1985).
Cultivars used here were divided into long (L) and short (S) hair types, taking 0.7 mm
as the dividing line.
K and Hr showed a negative correlation. Cultivars of WL type were less frequent
than WS, RL, or RS types. Cultivars of RL type frequently had the ph allele, while those
of WS type had the Ph allele. Thus, the cultivars tended to be divided into Ph WS and
ph RL types. These two types correspond to indica and japonica.
The distribution pattern of Z values is shown in Figure 2. Typical indicas had
positive Z values, and typical japonicas had negative Z values. The frequency of Z
values for all cultivars showed a continuous but bimodal distribution, indicating that
the cultivars used here showed a fair tendency to differentiate into indica and japonica
groups.
The pattern of association among nine genes and characters other than the three
discriminant characters is illustrated in Figure 3. Solid and dotted lines indicate
nonrandom associations significant at the 1% and 5% levels, respectively, which are
based on correlation coefficients (between quantitative characters), chi-square values
(between qualitative characters), or t values (between qualitative and quantitative
characters). Of the 36 possible combinations of genes and characters, 25 (69.4%)
showed nonrandom association.
Of these nine genes and characters, five genes (other than pericarp color, hull color,
awn, and Est-2 ) were used for varietal classification. The cultivars were classified by
allelic association at these five loci. Fourteen of 32 possible combinations (25=32,
Table 1) were found. Many cultivars were classified into a few representative
2. Frequency distribution of Z values showing indica-japonica variation among 200 cultivars.
48 Y .-I. Sato
3. Association of 9 genes and characters among 200 cultivars. Significance at the 1% (solid lines) and 5%
(dotted lines) levels.
Table 1. Classification of 200 cultivars into 32 genotypes based on nonrandom
association among 5 loci.
Allele at
Cultivars Coefficient of
C Pgi-2 Cat-1 Acp-1 Amp-2 (no.) estrangement a
C
C
C
C
C
C
C
C
C
C
C
C
C
C
1
1
1
1
1
2
1
1
2
1
2
1
1
2
2
2
1
2
1
1
1
1
1
1
1
2
1
1
Total
2
2
2
2
2
1
1
1
1
1
2
1
1
1
1
1
1
2
2
1
1
2
1
2
2
2
2
2
61
27
2
1
2
2
1
3
2
4
5
11
16
63
200
0
1
1
2
2
3
3
3
4
4
4
4
4
5
a Number of loci having alleles different from C Pgi-2 1 Cat-1 2 Acp-1 2 and Amp-2 1 type.
How was rice differentiated into indica and japonica? 49
genotypes, such as C Pgi-2 1 Cat-1 2 Acp-1 2 Amp-2 1 , c Pgi-2 1 Cat-1 2 Acp-1 2 Amp-2 1 , and
c Pgi-2 2 Cat-1 1 Acp-1 1 Amp-2 2 . The observed frequency was different from the
expected frequency calculated from random association of the genes. Genotype c Pgi-
2 2 Cat-1 1 Acp-1 1 Amp-2 2 was the most frequent. Its reverse genotype, C Pgi-2 1 Cat-1 2
Acp-1 2 Amp-2 1 and a similar one, c Pgi-2 1 Cat-1 2 Acp-1 2 Amp-2 1 , were also frequent.
This indicates that the cultivars used tended to be differentiated into c Pgi-2 2 Cat-1 1
Acp-1 1 Amp-2 2 and C Pgi-2 1 Cat-1 2 Acp-1 2 Amp-2 1 types.
The Z values of the cultivars belonging to these three genotypes are indicated in
Figure 2. All cultivars belonging to c Pgi-2 2 Cat-1 1 Acp-1 1 Amp-2 2 had positive Z
values, while those belonging to C Pgi-2 1 Cat-1 2 Acp-1 2 Amp-2 1 and c Pgi-2 1 Cat-1 2
Acp-1 2 Amp-2 1 had negative Z values. This means that the two most frequent
genotypes, C Pgi-2 1 Cat-1 2 Acp-1 2 Amp-2 1 and c Pgi-2 2 Cat-1 2 Acp-1 1 Amp-2 2 ,
correspond to indica and japonica, respectively.
Character and gene association in hybrid populations
The frequency distributions of Z values in the hybrid populations are shown in Figure
4. In the F 2 , the Z values showed a continuous and unimodal distribution, indicating that
indica-japonica differentiation did not occur. The mean of Z (1.27) was much higher
than zero, the value separating indicas and japonicas. This may be because the alleles
carried by indicas are largely dominant over japonica alleles. The mean of Z shifted
toward the negative from the F 2 to the F 5 because more segregants showed negative
values. The standard deviation of the Z value was 0.97 in the F 2 and became greater with
each generation, reaching 1.28 in the F 5 —still lower than that among the cultivars
(1.66). However, in the F 5 , the range of variation was as wide as that in the cultivars.
4. Z values in hybrid populations. Standard deviations in F 2 , F 3 , F 4 , and F 5 populations were 0.97, 1.06, 1.16,
and 1.28, respectively. Arrows indicate means.
50 Y.-I. Sato
The pattern of association among the nine genes and characters in the F 2 and F 5
populations is shown in Figure 5. Of 36 possible combinations, 4 in the F 2 and 6 in the
F 5 showed nonrandom associations. The associations found in the F 2 are probably due
to linkage (e.g., c, Est-2 1 , and Pgi-2 1 ).
In the F 5 population, three associations that did not appear in the F 2 were recovered.
They are the associations between black hull and Rc (red pericarp), between Rc and
Amp-2 2 , and between Est-2 1 and Acp-1 1 . In the associations between black hull and Rc,
and between Est-2 1 and Acp-1 1 , parental combinations of alleles were more frequent
than their recombinant types.
Discussion
Indicas and japonicas differ in genes and characters that are associated nonrandomly,
as indicated by many authors (Glaszmann 1987, Oka 1958, Sato et al 1986), who
concluded that no single gene representative of indica-japonica differentiation can be
pointed out. The genetic mechanisms responsible for such nonrandom association
among cultivars should be studied to elucidate factors causing indica-japonica differentiation.
The bimodality of Z values and the nonrandom association among nine genes and
characters recovered in the F 5 suggest that these gene and character associations could
be constructed against randomizing forces. Nonrandom associations between genes
and characters are not completely understood, even though considerable outcrossing
occurs. This may indicate that indicas and japonicas are genetically distant, because
recombinant or intermediate types decreased in the hybrids.
Nonrandom association between alleles is caused by various evolutionary forces
such as gametic selection, zygotic selection, random drift, linkage, and nonrandom
mating in higher plants (Hedrick et al 1978). Artificial selection also plays an important
role in cultivated species.
In the F 2 of the indica-japonica hybrid population, Z values showed continuous
variation. Moreover, the associations observed among the cultivars largely disappeared.
These facts indicate that the gene and character associations found among
5. Association among 9 genes and characters in F 2 and F 5 populations.
How was rice differentiated into indica and japonica? 51
cultivars are caused largely by natural or artificial selection. Random drift may not be
a causal factor of the indica-japonica differentiation, because it has been repeatedly
reported in Asian and African cultivars (e.g., De Kochko 1987).
Associations found in the F 2 population are likely to be caused by linkage (e.g., C ,
Est-2, and Pgi-2 ). However, three associations found in the F 5 but absent in the F 2 could
not be explained by linkage. In fact, in those three associations, the relevant loci are
known to be carried by different chromosomes ( Acp-1 and Est-2, Rc and Amp-2, and
Rc and Bh-a, Bh-b ). In addition, Ph and Rc, which are located on different chromosomes,
are also nonrandomly associated in the F 5 . Since the F 5 population was raised
by the single seed descent method, the effect of zygotic selection must have been
eliminated. Thus, these associations are considered to be caused by gametic selection.
Gametic selection is caused by various mechanisms. Hybrid sterility, in which
pollen having particular genes becomes abortive, is frequently observed in indica/
japonica crosses (Ikehashi and Araki 1986; Oka 1953, 1974; Oka and Doida 1962;
Yokoo 1984) and results in distorted segregation in the hybrids. Distorted segregation
also occurs in distant crosses by certation, which bring about differential fertilization
among normal pollen grains (e.g., Nakagahra 1972). However, it has been proven from
a computer simulation study that hybrid sterility and certation do not cause the
nonrandom associations found in hybrid populations (Nomura et al 1991). Perhaps the
gene and character associations found in the F 5 were due to gametic selection caused
by the differential fertilizing abilities of gametes with different genotypes. This trend
of gametic selection may act as an internal mechanism for indica-japonica differentiation.