Prospective use of Oryza longistaminata for rice breeding

Morphological types, fertility, and outcrossing rates were studied in a population
of 10 interspecific backcross progenies ( O. longistaminata/ O. sativa// O. sativa)
left under open pollination conditions. By segregation analysis at eight electrophoretic
loci, single-locus and multilocus estimates of the outcrossing rates were
calculated. In the first generation, 75% of the seeds came from outcrossing; this
rate decreased to 35% in the second generation, following pollen fertility restoration.
Outcrossing rates appeared primarily related to plant sterility and secondarily
to stigma length and exsertion. At the morphological level, an important diversity
of plant types was observed in the first generation, but plants were characterized
by various wild traits. The second generation spontaneously evolved toward a
more cultivated type, and transgressive segregants were observed for different
morphological traits. Allelic segregations at the F 1 level were normal, but the
second generation exhibited highly significant distortions. A loss of alleles coming
from the wild species was observed for 5 of the 8 loci and for all 10 families.
Oryza longistaminata is a wild species of rice that grows widely throughout intertropical
Africa. It covers a large range of ecological sites, from flooded plains to temporary
ponds, and propagates itself by developing vigorous rhizomes (Ghesquiere 1985). This
species is allogamous, with a self-incompatibility system, and shows the extreme
maximum values of stigma and anther length and number of pollen grains within the
Sativa species group (Oka and Morishima 1967).
This species shows significant diversity at the isozyme level (Ghesquiere 1988) and
appears to be among the most distant species from O. sativa within the Sativa group
(Second 1985). O. longistaminata has not intervened during the domestication of O.
sativa, nor in the latter’s diversification on the African continent since its introduction
there, because of the strong reproductive barrier that isolates the former from all other
species. This barrier is due to the action of two complementary lethal genes that cause
abortion of the embryo (Chu and Oka 1970a, Ghesquiere 1988). In spite of this barrier,
hybrid plants may be obtained, either by artificial crossing or, rarely, in seed sets
collected from wild plants along the borders of ricefields.



The spontaneous hybrids were called Obake plants by Chu and Oka (1970b). They
are characterized by different features shared with artificial hybrids: absence of
rhizomes, high tillering habit, photoperiod insensitivity, low male fertility, and
variable female fertility. When pollen fertility is not limiting, their inbred offspring are
highly depressed. The genetic structure of these plants has been extensively studied by
Ghesquiere (1988), who showed their possible origins from F 1 seeds to alternate
backcrosses of O. sativa and O. longistaminata.
The use of O. longistaminata in rice breeding has been envisaged for the introgression
of some of its allogamous characteristics into O. sativa to produce hybrid varieties
(Taillebois and Guimaraes 1987). O. longistaminata may also be useful for the
introgression of special disease resistance genes. This study considered the wild
species as a source of diversity in general, and as a source of allogamy to favor
spontaneous intermating in a hybrid population with a wide genetic base. This
experiment was conducted to show that it was possible to avoid the classical reduction
of variability of the backcross technique. The morphological diversity of a population
composed of backcross progenies between an Obake plant and 10 varieties of
cultivated rice was described for two generations. Outcrossing rates were estimated
and related to the fertility of the population. Intergenomic recombination was studied
at the isozyme level. The possibility of rapidly restoring a cultivated and fertile type,
and the efficiency of selection were estimated in a third trial.
Materials and methods
The maternal parent of all progenies was an Obake plant obtained from a seed collected
in North Cameroons. Its isozyme pattern was similar to that of a hybrid between O.
longistaminata and an indica variety of O. sativa. This plant was male sterile. It had
been pollinated by 10 upland rice varieties well adapted to West African conditions.
The 10 progenies (300 plants each) were observed in field trials during 2 generations.
To enhance natural intermating, the completely randomized trials were left under open
pollination conditions for two generations. Because the first (G 1 ) generation showed
significant variation in plant fertility, two trials were carried out in the second (G 2 )
generation. In one (the single seed descent [SSD] method), one seed of each G 1 plant
was cultivated; in the other (the bulk method), all the seeds of each family were mixed,
and a sample of 300 plants/family was cultivated. A third trial (G 3 ), composed of 20
self-pollinated progenies of G 2 plants chosen for their fertility, was studied. Each
progeny was composed of 60 plants. To compare the hybrids with their cultivated
parents in the G 1 and G 3 , the parental varieties were studied, but they did not participate
in the pollination. Measurements from single plants were taken to describe morphological
characters including plant height, tillering, flowering date, panicle architecture
(length, number of primary and secondary branches), pigmentation, awn development,
and seed shedding. Fertility was described by pollen fertility and total number of seeds
produced per plant. In the G 2 , the stigma and anther length and the rate of exserted
stigma were measured on plants whose progenies were used for estimation of
outcrossing rates.
82 Causse and Ghesquiere
Electrophoretic analyses, based on the technique described by Second and Trouslot
(1980), were performed on samples of each generation. Seven enzymatic systems were
studied, corresponding to nine polymorphic loci: Amp-1, Cat-1, Est-2, Est-5, Est-9,
Enp-1, Pgd-1, Pgi-1, and Sdh-1. Distinguishing the parental species origin of the alleles
was possible for five loci. Estimates of outcrossing rates in the G 1 were calculated from
the genotypic frequencies observed in the G 2 populations, and estimates of outcrossing
rates in the G 2 were calculated from the studies of open-pollinated progenies of G 2
single-plant arrays. Sixty-four progenies of six plants were analyzed. Two maximum
likelihood methods were used: one gives an estimate for each locus (method derived
from Brown and Allard 1970) and the other, based on the simultaneous analysis of
genotypes at different loci, gives a multilocus estimate (Shaw et al 1981). The maternal
genotypic frequencies were known in both cases, but pollen pool allele frequencies
were estimated on the same samples as outcrossing rates. Heterogeneity of the singlelocus
estimates was tested by chi-square analysis.
Results
The evolution of the population is first presented according to its reproductive behavior
and morphological traits. The extent of intergenomic recombination as shown at the
isozyme level is then presented. Finally, the implications of these results for the use of
O. longistuminata for plant breeding and genetic studies are discussed.
Reproductive behavior of interspecific progenies
The mean pollen fertility was very low in both generations: 18% in the G 1 and 37% in
the G 2 . Nevertheless, there was high variability between plants. In each generation, a
few fertile plants composed the efficient pollen pool. In the G 1 , 5% of the plants had
a pollen fertility above 60%; in the G 2 , 20% of the plants had a fertility slightly higher.
This low rate of pollinators is a source of drift. Nevertheless, the distributions of pollen
fertility among families were not statistically different. Seed production was also very
low: the mean was 20 seeds/plant in the G 1 and 47 in the G 2 . In both trials, 20% of the
plants produced no seed and therefore did not contribute to the next generation.
Although the SSD and bulk methods differed in seed contribution of G 1 plants to the
G 2 , the distribution of fertility of the plants in these two methods exhibited no
difference. The early selection, at low intensity, in the bulk method was inefficient in
increasing the fertility of the progeny.
In the third trial, selection of the most fertile plants in the G 2 was efficient, and strong
correlations were found between the pollen and seed fertility of G 2 selected plants and
the mean fertility of their offspring (0.68 and 0.75, respectively); yet large variability
was found in each progeny. These results show that it is possible to rapidly restore the
fertility of hybrid-derived plants.
The evolution of apparent outcrossing rates in the G 1 and G 2 is presented in Table
1. Significant differences between loci estimates were found in each generation.
Estimation of outcrossing rates is usually applied to natural populations with
homogeneous fertility. Differences between genotypic frequencies at the adult stage
Use of Oryza longistaminata for rice breeding 83
Table 1. Mean outcrossing rates (m) and standard deviation (SD) in G 1 (estimates based on
genotypic frequencies observed in 2 G 2 experiments) and G 2 (estimates based on genotypic
frequencies of 64 single G 2 plant progeny arrays). Estimates for each locus, means over
loci, and multilocus estimates.
Estimate G 1 G 2
from
locus G 2 :SSD (n=371) G 2 :bulk (n=304) Progenies (n=435)
m SD m SD m SD
Est-5
Enp-1
Amp-1
Cat-1
Sdh-1
Pgd-1
Mean over
Mean, multilocus
loci
1.165
1.124
0.886
0.472
0.563
0.891
(0.154)
(0.109)
(0.098)
(0.102)
(0.087)
(0.088)
0.82
0.74
0.564
1.040
0.662
0.184
0.551
0.579
(0.200)
(0.157)
(0.120)
(0.102)
(0.118)
(0.077)
0.63
0.54
0.492
0.519
0.340
0.582
0.240
0.031
(0.122)
(0.159)
(0.121)
(0.082)
(0.141)
(0.097)
0.35
0.35
and pollen frequencies might be a source of deviation in estimates, the existence of
selective factors usually being the cause of variations between estimates provided by
different loci (Kesseli and Jain 1985).
Here the artificial hybrid population was submitted to different selective pressures.
Despite certain aberrant values, the mean estimates look coherent. A comparison of
estimates provided by the two methods confirms this result, because the multilocus
estimate is supposedly less affected by variations between pollination and estimation
of genotypic frequencies (Shaw et al 1981). Actually, at least 75% of the plants
observed in the G 2 came from a detectable outcrossing event in the G 1 . In the G 2 , the
rate was reduced and involved 35% of the plants.
A relationship between inbreeding rate and pollen fertility was found at the level of
the individual: in the G 2 , plants whose pollen fertility was higher than 40% were
preferentially inbred. This relationship might also be extended to the succession of
generations. Lower values of the outcrossing rate estimates provided by the bulk
experiment, compared with the SSD experiment, reflect the importance of inbreeding
in the origin of bulk plants. In the G 3 progenies, outcrossing rates were estimated by
comparing seed production with obligate inbreeding and with open pollination. At this
level, the mean rate was 15%. The outcrossing rates were related to the stigma length
of plants ( r = 0.39) and to their exsertion rates, but were independent of anther length.
This result agrees with studies of floral characteristics influencing outcrossing in O.
sativa (Xu and Shen 1988).
Morphological changes along generations
In the G 1 and G 2 , considerable diversity was found for all the morphological traits
(Table 2). However, G 1 plants were characterized by very high and continuous tillering,
partly due to high general vegetative vigor and to the ability to emit new tillers from
84 Causse and Ghesquiere
Table 2. Means (m) and standard deviations (SD) of one of the 10 parental
varieties (P 6 ) and its progeny, in cross with an Obake plant, in G 1 and G 2 (SSD).
P 6 (n=10) G 1 (n=62) G 2 (SSD)
m SD m SD m SD
Character
Panicle length (cm)
Primary branches (no.)
Secondary branches (no.)
Panicle density a
Earliness (no. of days)
Tillering b
Height (cm)
25.5
15.5
34.0
94.2
24.1
92.5
2.19
1.05
0.93
4.71
0.27
3.62
9.30
7.1 6
26.8
11.4
31.5
102.7
37.1
87.7
2.76
3.71
2.44
10.35
0.67
7.15
18.00
13.16
23.8
9.9
29.8
2.36
110.9
24.7
75.9
4.70
3.74
21.72
1.25
12.39
15.20
15.41
a The ratio of secondary to primary branches. b Number of panicles produced in 4 wk after the
beginning of flowering.
buds located on aerial nodes. This capacity, observed in every family, allowed the
overlapping of flowering periods of plants with different growth durations. Panicles
were usually longer but with fewer branches than those of the cultivated parents,
although the ratio of secondary to primary branches appeared higher in G 1 plants than
in the cultivated parental varieties. O. longistaminata plants and the mother Obake
plant, observed in irrigated conditions, showed long panicles with few secondary
branches. Other wild traits characterized hybrid progenies: they were frequently
pigmented (60% of G 1 plants showed collar, spikelet, or stigma pigmentation) and long
awned, and, where measurement was possible, shedding was high in almost every
plant. Rhizomes never developed; this trait appeared to have been totally eliminated
since the first hybridization.
G 2 plants evolved toward a more cultivated type. A reduction in the rates of
pigmented plants and awned spikelets was observed with lower tillering habit and
reduced perenniality. The genetic load of O. longistaminata, a wild and allogamous
species, was expressed through weakness and many panicle or spikelet abnormalities.
At the same time, an increase of total variance was found for every trait, while the
between-family variances decreased (Table 3). Interspecific hybridization does not
seem to induce high hereditary modification of quantitative traits. Analysis of the
regressions between the cultivated parents of the population and their progenies in the
G 1 and G 2 shows intermediate values for the majority of characters, all significantly
different from zero, except for height (Table 3). Higher coefficients were observed in
regressions between G 2 plants and their inbred progenies, indicating the expected
efficiency of selection within these progenies. Though the choice of G 3 mother plants
had been based only on the fertility of the hybrids, comparison of the G 3 progenies with
four cultivated rice varieties showed the existence of transgressive segregants for all
traits studied, particularly for early developmental characteristics and panicle architecture.
In the G 1 and G 2 , the majority of the plants were arrested, and seed shedding was
important. Segregation for these traits was analyzed in the G 3 progenies in relation to
the parental phenotypes. These two characters showed complex segregation and
Use of Oryza longistaminata for rice breeding 85
Table 3. Comparison of 1) ratios of between-family variance (Vb) to total
variance (Vt) for the 10 parental varieties (Pi), and the corresponding families
in G 1 and G 2 (based on 10 families of 62 individuals), and 2) coefficients of
parent-offspring regressions for different generation associations. a
Character
Panicle length
Primary branches
Secondary branches
Earliness
Tillering
Height
Vb/Vt Regression coefficients
Pi G 1 G 2 Pi/G 1 Pi/G 2 G 2 /G 3
0.611
0.797
0.580
0.762
0.836
0.334
0.157
0.176
0.099
0.271
0.123
0.095
0.052
0.010
0.023
0.064
0.045
0.009
0.732
0.378
0.301
0.337
0.262
0.018
0.533
0.2220
0.344
0.502
0.129
0
0.674
0.765
0.462
0.679
0.117
0.330
a For parent to G 1 (Pi/G 1 ) and G 2 (Pi/G 2 ), the regressions are based on 10 families of 62 individuals;
for G 2 to G 3 (G 2 /G 3 ), the regressions are based on 20 progenies of 30 individuals.
appeared independent of each other. Some fertile and nonshedding offspring were
observed. Segregation of seed shedding is independent of plant fertility. These results
are consistent with those of Morishima (1985), who suggested that the “domestication”
traits had been acquired by the accumulation of genes spread throughout the genome.
Intergenomic recombination
Linkage studies were carried out on 2 G 1 and 10 G 3 progenies. Except for Enp-1 and
Cat-1, which were found tightly linked (with 0.04 and 0.13% recombinants in the 2
progenies studied), all the other markers were independent. This result is consistent
with previous studies of isozyme location (Pham et al 1990). Knowing the parental
genotypes, allelic frequencies were compared with expected ones (Table 4). The
gametic segregation of the Obake plant, tested through G 1 genotypes, appeared
significantly skewed for only one ( Amp-1 ) of the seven loci studied, although others
also showed slight losses of the alleles coming from the wild species; Sdh-1 was the
only exception. In G 2 SSD, significant losses of wild alleles were found for five loci
( Amp-1, Est-9, Enp-1, Pgd-1, Sdh-1 ). The importance of the loss varied between
families and between loci. Differences between bulk and SSD frequencies were a
consequence of differences in the contribution of G 1 plants to the next segregation. In
bulk, losses were of the same or had increased intensity for all families for loci Amp-
1, Enp-1, Pgd-1, and Pgi-1 ; they differed from one family to another for Sdh-1 and Est-
9. Study of open-pollinated progenies of G 2 plants (used mainly to estimate outcrossing
rates) allowed us to follow this evolution. Losses seemed to become stable for loci Pgi-
1 and Enp-1 but still increased for Est-5, Amp-1, and Sdh-1.
Segregation was also studied in inbred G 3 progenies, which presented different
parental genotypes. Among 43 segregation patterns studied, 11 showed significant
distortions, but their intensity and direction varied between loci and progenies. Amp-
1 was the only locus to show high loss of the wild allele in both progenies where it
segregated. Est-5 and Cat-1 were studied in five progenies; they showed loss of the wild
86 Causse and Ghesquiere
Table 4. Evolution of the frequencies of the alleles coming from O. longistaminata
in the backcross population over generations. a
Locus G 1 G 2 -SSD G 2 -bulk G 3 -op
Est-5 0.200** 0.140** 0.090*** 0.095***
Enp-1 0.230 0.163** 0.1 35*** 0.135***
Amp-1 0.200** 0.140** 0.140** 0.125***
Sdh-1 0.270 0.195* 0.230 0.140**
Pgi-1 0.209 0.186* 0.090*** 0.095***
a Level of significance of the test of comparison between observed and expected frequency =
0.25. Estimates are based on the genotypes of 170 plants of the G 1 , 371 and 304 of the G 2
observed in SSD (G 2 -SSD) or bulk (G 2 -bulk), and 435 open-pollinated offspring of G 2 plants (G 3 -
op).
allele in one, an excess in another, and no distortion in three others. Est-2, Pgi-1, and
Sdh-1 studied in four, six, and five progenies, respectively, showed no distortion.
Contrary to expectations from the comparison of allelic frequencies in SSD and bulk,
a systematic analysis of the relationship between isozyme genotype and fertility
showed no significance. Losses of wild alleles could be attributed not only to
differences in fertility between genotypes but also to other sources of deviations, linked
or not to the genetic load of the wild species, such as losses at germination, weakness,
or selective assortment of gametes.
Discussion and conclusions
This experiment showed that it is possible to obtain spontaneous intermating in a wide
experimental population despite a high level of sterility. The mixing of two distant
species under two mating systems led to highly sterile plants; few fertile plants
composed the efficient pollen pool in each generation. Long and well-exserted stigmas,
contributed by the wild species, tended to enhance outcrossing rates. Perenniality,
expressed by continuous tillering, also favored intermating between plants of different
families. Because of this important outcrossing rate in the G 1 , variability remained high
in the G 2 . However, the parallel evolution of pollen fertility and inbreeding showed the
limits of the intermating system. Nevertheless, introgression of the long stigma in a
male sterile variety should lead to increased outcrossing rates necessary for high
production of hybrid seed.
Spontaneous evolution of the population toward a more cultivated type was found
at the morphological and isozyme level, and different origins were proposed for that
evolution. Evidence for genetic exchanges between cultivated rices and O. longistaminata
was suggested at the isozyme level, but because of its reproductive barrier, this
species did not play a major role in the diversification of cultivated rices in Africa.
Nevertheless, a very high morphological diversity could be expected in the offspring
of such crosses. With distant phenotypes and complementary growth habits, introgressive
hybridization may be useful for rice breeding. Though plants in the first generation
were not attractive from a plant breeding point of view, it was possible to restore
Use of Oryza longistaminata for rice breeding 87
fertility and to eliminate unfavorable traits in a few generations. Interesting transgressive
segregants were observed in G 3 progenies. Without major variation in the
heritability of quantitative traits, selection among and within the progenies might be
efficient. Even if introgressive lines derived from this cross are not directly exploitable
as cultivated varieties, they might be an interesting source of new diversity for rice
breeding.
New technologies, such as restriction fragment length polymorphism (RFLP)
marker-assisted selection, may facilitate the elimination of unfavorable wild traits.
Shedding and awning are governed by a few genes spread throughout the genome,
which should be easily located by RFLP studies. The detection and location of
quantitative trait loci involved in other morphological traits that distinguish rice
species might also help greatly in selecting hybrid derivatives. In an interspecific
population showing a strong linkage disequilibrium, selection by target markers should
be easy. An RFLP map of the rice genome is already available (McCouch et al 1988)
and has been transferred to an interspecific backcross population involving O.
longistaminata (Tanksley et al 1991), which will allow testing such a hypothesis.