Agronomic Performance and Selection of Doubled- Haploid Rice Lines for Rainfed Lowland Paddy Field

Rainfed lowland rice cultivation is an alternative to increase national rice production. Breeding of high yielding rice varieties suitable for rainfed lowland condition can be accelerated by using doubledhaploid (DH) as genetic materials. This study aimed at obtaining information on the agronomic performance including yields in several DH rice lines and selecting DH lines suitable for rainfed lowland paddy field. The experimental design used was a Randomized Complete Block Design with 3 replications. The treatment was thirty DH lines and 4 check varieties namely RJ31 Ciherang, RJ32 Inpari 18, RJ33 Inpari 40, and RJ43 Inpari 41. The results showed that there were variability in all agronomic performances, i.e., plant height, number of tillers, days to heading and to harvest, panicle length, number of filled and empty grains, 1000-grain weight and grain yield. The DH lines, namely RJ19 DR8-43-3-1 and RJ25 DR10-141-1, gave the same productivity as 4 check varieties. Index selection showed that twelve DH lines with medium number of productive tillers, early maturing, and productivity of more than 4.40 tons.ha-1 were selected for further evaluation.


Introduction
Rice (Oryza sativa L.) is the main staple food for most Indonesian and Asian people. Rice has been cultivated for a long time in Indonesia. Preserved ancient botanical evidence in the form of rice phytoliths has confirmed that people farmed domesticated rice in the interior of Sulawesi Island, Indonesia, by at least 3,500 years ago (Deng et al., 2020). Recently, national rice consumption reaches 98.40 kg.capita -1 .
The problems of productivity, drought tolerance, and short duration rice, among others, can be overcome by plant breeding activities to get the desired lines. Obtaining these lines can be done with conventional breeding or anther culture. According to Dewi and Purwoko (2001;2011) breeding carried out conventionally takes a very long time from 8 to10 generations. Shortening of pure line development time can be done by utilizing the haploid system through anther culture technique. It is expected that the process of obtaining pure lines as fully homozygous lines (doubled-haploid /DH) is accelerated by only one to two generations. Thus, it can increase efficiency of the selection process and accelerate the acquisition of varieties for release as well as saving costs, time, and labor (Dewi and Purwoko, 2012;Mishra and Rao, 2016;. New improved varieties can be determined by the agronomic performance (phenotypic appearance) of a line. The quantitative character of genotypes can vary from one environment to another. A variety that is consistent with high yields in all environments is generally the desired genotype in breeding programs (Pabendon and Takdir, 2000). Plant breeding techniques are not only applied to the development of high yielding varieties but also the ability of varieties to adapt in various environments (Mulusew et al., 2009;Akter, 2014). Estimation of plant adaptation in various locations and the magnitude of the influence of interactions in the typology of land: biophysical (climate specifically), edaphic and biotic diversity (Sitaresmi et al., 2012) in a good agroecological environment is one technique for obtaining superior varieties from several rice lines tested. Munarso et al. (2010) stated that to identify stable and high yielding genotypes, evaluation in several target production environments should be conducted.
In previous research, Dewi et al. (2017) obtained 275 DH rice lines by using anther culture of F1 derived from crossing of parents that had high yield and drought tolerance characters. Those DH lines were then evaluated based on their agronomic characters. The results showed high genetic variability as well as broad sense heritability (more than 90%) for all variables tested. Anther culture could generate high genetic variability for further selection (Syafii et al., 2018). In this study 30 DH lines were evaluated for their agronomic performance. Then, potential DH lines suitable for rainfed lowland paddy fields were selected.

Material and Methods
This research was carried out at Sukamandi Rice Field Experiment Station, West Java, The Indonesian Center for Rice Research, from May to September 2019.

Planting and Harvesting
For each line, seeds were sowed in two 0.5 m long grooves with a distance between lines 10 cm. Land preparation involved clearing, ploughing, and harrowing. Transplanting was done by planting 18 day-old rice seedlings, 2 to 3 seedlings for each hill. Plant spacing used was 25 cm x 25 cm, so that in each plot there were 8 rows with 20 hills for each row.
Fertilization was done by giving NPK in the form of Urea (250 kg.ha -1 ), KCl (150 kg.ha -1 ), and SP36 (150 kg.ha -1 ). Fertilizers were given in three stages. The first fertilization was carried out at transplanting by giving Urea 83.3 g.plot -1 , KCl 150 g.plot -1 , and SP36 150 g.plot -1 . The second fertilization was carried out at 4 weeks after planting (WAP) by giving urea 83.3 g.plot -1 . The last fertilization was carried out at 8 WAP or during flowering stage by giving Urea 83.3 g.plot -1 . Maintenance includes weeding and controlling plant pests and diseases. Weeding was done intensively when the plant was in the vegetative and reproductive phases. Pest and disease control was carried out by mechanical and chemical methods. In mechanical method, pest was killed directly or the environment was made unsuitable for it. For example, traps for pest animals and insects; or barriers such as screens or fences to keep animals and insects out, while in chemical control, chemical pesticides are applied to protect plants from pests and diseases. Harvesting was carried out when 90% of rice panicles in one plot turned yellow. Seeds were harvested to estimate grain yield per plot. The grains were sun-dried for approximately 3 to 4 days to reach ± 14% moisture content as measured by grain moisture tester DMC550 and later converted to dry grain yield per hectare (ton.ha -1 ).

Experimental Design and Data Collection
The experimental design used was a completely randomized complete block design (RCBD) in the form of genotypes as a treatment (Table 1) which was repeated three times. The experimental unit was a 2 m x 5 m rice plot.
Observations and measurements of the yield components were carried out on 3 hills for all lines tested. The agronomic characters observed according to Standard Evaluation System for Rice from IRRI (2013) were as follows: 1. Vegetative plant height (cm) was measured from ground level to the tip of the highest leaf. This character was observed at 45 days after planting.
2. Generative plant height (cm) was measured from ground level to the longest panicle tip. This character was observed before harvest. 3. Number of productive tillers (tillers/hill) was measured before harvest by counting tillers that produce panicles. 4. The days to heading was calculated from the days of sowing until 50% of rice panicle in a plot was fully visible. 5. The days to harvest was calculated from the days of sowing to the days when 90% of rice panicles in one plot turned yellow. 6. Panicle length (cm) was measured from the panicle neck to the panicle tip of each panicle. 7. The number of filled grains per panicle (grains) was observed by counting the number of filled 8. grains in each panicle. 9. The number of empty grains per panicle (grains) was observed by counting the number of empty grains in each panicle. 10. Total number of grains per panicle (grains) was observed by counting the number of grains in 11. each panicle. 12. Percentage of unfilled grains per panicle (%), was obtained by comparing the number of unfilled grains per panicle to the number of grains per panicle then multiplying by one hundred. 13. Percentage of filled grains per panicle (%), was obtained by comparing the number of filled grains per panicle to the number of grains per panicle then multiplying by one hundred. 14. Weight of 1,000 grains (g), was done by weighing 1,000 filled grains with ± 14% moisture content. 15. Productivity (tons.ha -1 ) was calculated using the formula: (160,000 hills.ha -1 / number of hills harvested per plot ) x (grain yield per plot(kg)/ 1,000)

Data Analyses
Data obtained was analyzed by the normality test, if the tested trait distributed normally then continued with the analysis of variance. Furthermore, if the F test was significant, then Duncan's multiple range test was performed (Gomez and Gomez, 1984). Selection index was used to select genotypes suitable for further yield trials in rainfed lowland environment. Several agronomic important traits representing yield and yield components were chosen simultaneously and economic weightage was given to the phenotypic values of each trait in such a way that expected gain in aggregate genotypic value would be maximized (Ramos et al., 2014;Gazal et al., 2017). Determination of the selection index were conducted based on Falconer and Mackay (1996) as follows: where: I is the selection index bn is the weight of the variable-n Xn is a standardized phenotype value (z) for variable-n

� � � � �̅
x is the means of each genotype is the means of the variable s is the standard deviation of the variable Then, the I values were ranked and used to select the best lines. The analysis of variance and the construction of the weighted selection index were carried out using SAS 9.0 and STAR programs.

Agronomic Characters of Doubled-Haploid Lines
The agronomic performance of DH rice lines showed different responses. This result can be seen in the recapitulation of variance observed in 12 characters (Table 2). Genotypes have a significant effect on all agronomic traits observed, except for the number of grains per panicle and the percentage of filled grains per panicle. The analysis of variance in Table 2 shows the character such as days to harvest has the lowest coefficient of variation (CV) value, while the number of empty grains has the highest CV value. The CV is the ratio of the standard deviation to the mean. When we are presented with estimated values, the CV relates the standard deviation of the estimate to the value of this estimate. However, the magnitude of the CV value depends on the experimental, plant, and observed factors. The lower the value of the CV, the more precise the estimate, so that it will reduce the error of the experiment (Gomez and Gomez, 1984).
Plant height character was measured at vegetative and reproductive stages. The average vegetative plant height of 34 rice genotypes ranged from 65.4-115.7 cm, while for reproductive plant height ranged from 75.6-141.9 cm (Table 3). Vegetative plant height will increase until the plant reached the reproductive stage. According to Rachmawati and Retnaningrum (2013), plant height increase is influenced by the availability of water and nutrients. Plant height at reproductive stage is one of the important criteria in selection process because it is related to the ease of maintenance and harvesting process. Tall rice plants (> 125 cm) are very easy to lodge and lodging can cause a decrease in grain yield, while the short one (< 80 cm) is relatively difficult for farmers to harvest. Therefore, plant height at reproductive stage is also an important character to farmers' acceptance of new varieties (Dewi et al., 2015). In general, according to Akbar et al. (2018) intermediate plant height (90-124 cm) was categorized as a good agronomic character for rainfed rice.
The average number of productive tillers ranged from 11.6 to 35.2 tillers per hill (Table 3). From this experiment, there is one DH line, RJ23DR9-58-1-1, that has a desirable number of productive tillers (35 tillers per hill), 6 DH lines have an acceptable number of productive tillers (20-25 tiller per hill), and the rest have medium number of productive tillers (10-19 tillers per hill). All DH lines, except for RJ23DR9-58-1-1 line, have non-significantly different number of productive tillers with the check varieties for rainfed lowland, i.e. Inpari 18, Inpari 40 and Inpari 41. The number of tillers has been reported to have a positive association with plant biomass and economic yields in rice (Wang et al., 2017). Besides the genetic and environmental conditions, the number of rice tillers is strongly influenced by age of seedling at transplanting time and plant spacing which will also determine its population density (Donggulo et al., 2017). Especially for rice, the number of productive tillers often became the main character in the selection process because it directly influences rice yields through the number of panicles, the number of grains, and filled grains per panicle (Fukushima, 2019). However, the large number of productive tillers can result in the nonsynchronized panicle maturity, thus causing the quality of rice to decline (Abdullah et al., 2008).   recommended to be planted in rainfed rice fields. This is in accordance with the conditions of rainfed rice fields where the source of water for irrigation is limited from rainfall which is difficult to predict. Early maturity rice is also preferred by farmers at irrigated lowland rice fields because it can increase cropping index from two to three times a year (Fatimah et al., 2014).
The number of filled grains per panicle is one of the selection criteria to obtain lines with high yields. The results of the analysis showed the average number of filled grains per panicle of DH lines ranged from 102.7-189.2 grains, while the number of unfilled grain ranged from 27.8 -106.3 grains (Table 5). In this experiment, twenty-three DH lines has non-significantly different number of filled grains compared to the four check varieties. The number of filled and unfilled grains is influenced by genetic and environmental factors. Spikelet fertility and seed yield per panicle were severely reduced by extreme temperature in the 14 days period before anthesis (Eixarch and Ellis, 2015). Too low temperature affected spikelet fertility (Zeng et al., 2017), while too high temperature will inhibit the development of pollen grains in the anther resulted in sterile spikelets and unfilled grains, thus decrease the number of filled grains (Khamid, 2016). Other factors that influence the amount of filled and unfilled grains include the level of pests and diseases attacks during planting season.
The number of grains per panicle ranged between 162.7 -274.7 grains (  Abdullah et al. (2008) stated that the high number of grains per panicle will extend maturation phase, thus it will increase the number of empty grains due to imbalance of source and sink, especially in new plant type (NPT) of rice.
The results of the analysis of the percentage of filled grain are presented in Table 6. Check variety RJ34 Inpari 41 has the highest average percentage of filled grain of 86.9%. There are 12 DH lines that have percentage of filled grain not significantly different from RJ34 Inpari 41. The grain yield is strongly influenced by several factors, namely genetic and environmental factors. Determination of the variety used, the choice of growing environment, the amount of light intensity, and the presence of pest and disease determine how much yield to be harvested (Wahyuni et al., 2013).
Productivity of DH lines ranges from 2.10 to 7.12 tons.ha -1 (Table 6). RJ7 DR7-43-1-5 has the lowest productivity while the check variety RJ31 Ciherang has the highest productivity. There are two DH lines, i.e., RJ19 DR8-43-3-1 (7.12 tons.ha -1 ) and RJ25 DR10-14-1-1 (6.45 tons.ha -1 ) which has the same productivity as the four check varieties. Mahmud and Purnomo (2014) stated that productivity is closely related to the yield components such as the number of productive tillers per hill, the number of grains per panicle, the percentage of filled grains, and the 1,000-grain weight.

Correlation Analysis for Determining Good Agronomic Performance
The relationship between agronomic traits and yield is known from the correlation analysis values presented in Table 7. The generative plant height correlates positively and very significantly with the vegetative plant height (r=0.85, P<0.01). The high plant height at reproductive stage is a result from high plant height at vegetative stage, which is not desirable (Dewi et al., 2015). The heading character correlates positively and very significantly with days to harvest (r=0.52, P<0.01). Faozi et al. (2010) stated that days to harvest was determined by the speed of heading, because in general rice plants have different duration of vegetative period, while the grain filling period is relatively similar.
The number of grain was positively and very significantly correlated to the character of the number of filled grains per panicle (r=0.71, P<0.01). The results of this analysis are in line with the results of Kartina et al. (2016) which stated that an increase in the total number of grains per panicle is accompanied by an increase in the overall number of filled grain. Increasing the number of filled grains per panicle will be followed by the addition of grain yield per hill. Moreover, grain yield per hectar or productivity positively correlates with the number of vegetative tillers, the number of filled grains per panicle, and the percentage of filled grain, but negatively correlated with the number of unfilled grains and the percentage of unfilled grains. The large correlation coefficient shows a close relationship between the characters observed (Prabowo et al., 2014). In this experiment a close relationship appeared between vegetative plant height and generative plant height, days to 50% flowering and days to harvest, number of unfilled  grain to percentage of filled grain, number of unfilled grain to percentage of unfilled grain, % number of filled grain to grain yield, percentage of unfilled grain to grain yield per hectar. Selection of doubled-haploid rice lines with superior agronomic character and high yield can be done through indirect selection using characters correlated to productivity (Aryana, 2009). The selection method used is weighted selection index which involves productivity (GY) and selected variables that have a significant effect to grain yield from analysis correlation, i.e., number of filled grains (NFG) and percentage of filled grain (%FG) and characters suitable for rainfed lowland paddy field, i.e. number of productive tiller (NPT) and days to harvest (DH). Before establishing the selection index model, the selected traits were given a weight of 1 to 5 to maximize the model as suggested by Sabouri et al. (2008). Therefore, the model of selection index (I) was formulated as follow: I = (5*GY) + (2*NPT) + (2* NFG) + (1*%FG) -(1*DH).

Selection Index to Obtain Superior Lines
The high weighting of the character of productivity and agronomic characters related to productivity in that formulation was based on the economic value (Ramos et al., 2014;Gazal et al., 2017). In addition, positive and negative signs indicated the direction of selection, i.e. increased number of filled grain and decreased days to harvest, respectively.

Conclusion
The DH lines tested in this study have variability in agronomic traits. Number of tillers at vegetative stage, number of grains and filled grain per panicle, and percentage of filled grain positively and significantly correlated to grain yield or productivity, while plant height at vegetative and generative stages, days to harvest, panicle length, and number of unfilled grains negatively correlated to grain yield. There were twelve DH lines that have potential to be tested further in yield trials program for rainfed lowland paddy field. The selected DH lines have good agronomic characters, especially in number of productive tiller, early maturing, and productivity of more than 4.40 tons.ha -1 .