WA-CMS-based iso-cytoplasmic restorers derived from commercial rice hybrids reveal distinct population structure and genetic divergence towards restorer diversification

Abstract

One hundred diverse iso-cytoplasmic restorer (ICR) lines carrying WA cytoplasm indicated significant but moderate variability for agro-morphological traits as well as for the microsatellite-based allele patterns. There were two major groups of ICRs based on agro-morphological clustering. Simple sequence repeat (SSR) markers identified allelic variants with an average of 2.48 alleles per locus and the gene diversity (GD) ranged from 0.02 to 0.62 at different loci. ICR lines showed a genetic structure involving two sub-populations, POP1 and POP2. Both the subpopulations had the presence of admixture lines. Nearest ancestry-based grouping of ICRs by neighbour-joining (NJ) method showed near similar grouping as that of sub-population division. The POP2 was the largest group but with fewer admixed lines. POP1 was more distinct than POP2. Since the hybrid parents of the ICRs had limited diversity on maternal lineage, paternal lineage was concluded as the major contributor to the observed divergence and population differentiation. ICRs developed from certain hybrids were more genetically distinct than other hybrids. Even with the moderate variability, ICRs could be considered as a potential source of fertility restoration in hybrid development because of their distinct population structure and the full complement of restorer genes they contained. ICR lines with high per se performance can be utilized in hybrid rice development by estimating their combining ability.

Electronic supplementary material

The online version of this article (10.1007/s13205-019-1824-3) contains supplementary material, which is available to authorized users.

Introduction

Rice is the most important crop in India, which plays a critical role in ensuring food security. It is the staple food of more than 60% of the population, being grown on more than 44 million hectares, the highest area ever occupied by a single crop in the country. Rice contributes to 47% of cereal production in India (agricoop.nic.in), as a result of an unprecedented growth due to the adoption of semi-dwarf high yielding varieties combined with the adoption of intensive input-based management practices. Rice production in India registered an increase from 20.58 million tonnes of milled rice in 1950–51 to 111.01 million tonnes during 2017–18 (eands.dacnet.nic.in). The corresponding increase in productivity was 668–2500 kg/ha during the same period. However, the genetic gain during the last three decades has been approaching a plateau, which is evident from the decrease in the annual compounded growth rate from 3.8% in 1971–80 to 2.0% during 2000–2010 (agricoop.nic.in).

Hybrid rice technology is a potential, viable, sustainable and cost-effective options to break the yield ceiling in rice (Sheeba et al. 2009). Rice hybrids realized 15–20% higher grain yield over the best semi-dwarf inbred varieties (Virmani et al. 2003). Hybrid rice technology in India is primarily based on the indica gene pool, which uses the most popular wild abortive-cytoplasmic genetic male sterility system (WA-CMS) for hybrid seed production. WA-CMS is a three-line system that requires a male sterile (A) line, its iso-genic maintainer (B) line and restorer (R) line (Subbaiyan et al. 2012). Since heterosis is fundamental to the yield superiority of hybrids, maintaining adequate diversity among the parental lines is essential for hybrid development. Thus, rice hybrid breeding using WA-CMS system requires the development of diverse restorer lines which can combine well with different male sterile lines. This will increase the options of using different cross combinations for the generation of superior hybrids, which in turn can prevent genetic vulnerability due to the use of limited genetic resources and thus ensuring a strong hybrid rice breeding program.

Fertility restoration in rice CMS systems is facilitated by nuclear-cytoplasmic interactions, that can display varying fertility levels. Iso-cytoplasmic lines are those carrying the same cytoplasm in which interactions between the cytoplasm and the fertility restoring (Rf) genes are maintained at a similar level. Iso-cytoplasmic restorer (ICR) lines derived from WA-CMS-based rice hybrids with more than 85% spikelet and pollen fertility (Kumar et al. 2017a) have a unique feature that they all carry WA cytoplasm (Kumar et al. 2017b). They offer the advantage of capturing the differences in fertility restoration due to the interaction of restorer genes and cytoplasm in these lines, thereby eliminating the need to screen for complete and partial fertility/sterility in derived test crosses (Kumar et al. 2017b). There has been limited effort to assess and identify potential ICR lines with good combining ability that could be used in hybrid rice development. In the present study, a set of 100 ICR lines developed from elite rice hybrids were characterized for agro-morphological traits as well as for genetic variation based on SSR markers.

Materials and methods

Plant materials

One hundred ICR lines derived from 25 elite commercial rice hybrids using pedigree breeding in F₇ generation by imposing selection for yield and panicle exsertion throughout successive generations of selfing (Kumar et al. 2017b) and maintained at Division of Genetics, ICAR–Indian Agricultural Research Institute (ICAR–IARI), New Delhi were used in the present study (Table 4). All the commercial hybrids were based on WA-CMS system, and, therefore, all the ICR lines carried WA cytoplasm. These ICRs were shortlisted from a pool of 390 lines by providing equal representation for each hybrid, and by selecting the top four ICRs per hybrid, after evaluation for yield and panicle exsertion.

Table 4.

Sub-population wise distribution of iso-cytoplasmic restorers based on hybrid origin

Parent hybrida	Parentageb	ICR linesc	Sub-population representedd
DRRH 2	IR 68897A/DR 714-1-2R	PRR 300–PRR 303	POP2 (4)
DRRH 3	APMS 6A/RPHR 1005	PRR 304–PRR 307	POP2 (4)
PSD 3	UPRI 95-17A/UPRI 93-287R	PRR 308–PRR 311	POP2 (4)
PRH 10	Pusa 6A/PRR78	PRR 312–PRR 315	POP2 (4)
CORH 3	TNAU CMS 2A/CB87R	PRR 316–PRR 319	POP2 (4)
Sahyadri 1	IR 58025A/BR 827-35-3-1-1-1R	PRR 320–PRR 323	POP2 (3), POP1 (1)
Sahyadri 2	IR 58025A/KJTR 2	PRR 324–PRR 327	POP2 (4)
Sahyadri 3	IR 58025A/KJTR 3	PRR 328–PRR 331	POP2 (4)
Sahyadri 4	IR 58025A/KJTR 4	PRR 332–PRR 335	POP2 (4)
GK 5003*	GK5003A/GK5003R	PRR 336–PRR 339	POP2 (4)
US 312*	F1/M66	PRR 340–PRR 343	POP2 (4)
NK 5251*	RC5076A/RC5056R	PRR 344–PRR 347	POP2 (4)
INDAM 200-017 *	IAHS24A/IASN707R	PRR 348–PRR 351	POP1 (2), POP2 (2)
DRH 775*	DRH775A/DRH775R	PRR 352–PRR 355	POP1 (3) POP2 (1)
PA 6129*	6CO2/6M10	PRR 356–PRR 359	POP2 (4)
PA 6201*	CO 2/MO 1	PRR 360–PRR 363	POP1 (4)
PA 6444*	IR62871-138-5	PRR 364–PRR 367	POP1 (4)
PHB 71*	–	PRR 368–PRR 371	POP1 (4)
Indira Sona	IR 58025A/R-710-437-1-1	PRR 372–PRR 375	POP1 (3), POP2 (1)
Suruchi 5401*	PMS79/PR319	PRR 376–PRR 379	POP1 (4)
JRH 8	IR 68897A/NPT29	PRR 380–PRR 383	POP1 (4)
JKRH 401*	RV2A/RV44R	PRR 384–PRR 387	POP1 (4)
PAC 835*	835A/835R	PRR 388–PRR 391	POP1 (4)
PAC 837*	837A/837R	PRR 392–PRR 395	POP1 (4)
KRH 2	IR 58025A/KMR 3R	PRR 396–PRR 399	POP1 (4)

ICR iso-cytoplasmic restorer

^aParental hybrids are either from the public sector or private sector institutions; public sector institutions are government funded, private sector institutions are private seed research laboratories/companies; Private sector hybrids are suffixed by * mark

^bParents of private sector hybrids are either codenamed or remain undisclosed

^cICRs are serially numbered with prefix ‘PRR’ from PRR 300 to PRR 399

^dThe numbers in brackets indicate the number of lines grouped in a particular sub-population

Screening for morphological traits

The lines were raised in an augmented randomized block design with four blocks and eight checks. The field experiments were carried out at two locations, in Delhi for two successive years (2014 and 2015) during Kharif season and at Aduthurai during the Rabi season of 2014–15, by adopting standard agronomic practices. The maintainers of four male sterile lines (IR 79156B, IR 58025B, Pusa 6B and RTN 12B), three promising restorers namely, PRR 78, RPHR 1005R and DRR 714R and one improved variety, Pusa Basmati 1609 were used as checks.

Data was recorded on five healthy plants from the middle row for each of the lines for various agro-morphological traits namely days to 50% flowering, plant height, panicle length, tiller number, grains per panicle, pollen fertility (%), spikelet fertility (%), test weight and yield per plant as per Xiao et al. (1998). Pollen fertility was determined by light microscopy using I-KI staining method following the protocol given in the Standard Evaluation System for rice (IRRI 2013). Combined ANOVA was performed to know the significant differences among the restorer lines.

Analysis of genetic divergence

The genetic divergence of the restorer assembly was estimated using both morphological (phenotypic) as well as genetic data (genotypic). For agro-morphologic data, the squared Euclidean distances between genotypes were computed and used for clustering using neighbour joining algorithm. For estimating genotypic diversity, a panel of 50 standard SSR markers suggested by the Generation Challenge Programme (GCP) of the Consultative Group for International Agricultural Research was used. The details of these markers are available in the Gramene database Release #39 (http://archive.gramene.org/markers/microsat/50_ssr.html). Originally shortlisted for genotyping with all AA genome of Oryza (www.gramene.org), these markers could effectively resolve population structure of rice. It has been reported that even a subset of 35–36 markers were sufficient enough to detect the population structure of rice (Ali et al. 2011; Roy et al. 2016). The lines were grown in pot trays with 12 plants in each well for DNA extraction and marker analyses. Fresh tissues from green, young and healthy plants were collected 10 days after germination and used for DNA extraction using standard CTAB procedure (Murray and Thompson 1980). The pure DNA was diluted after quality check and used for polymerase chain reaction using the SSR markers. The amplified DNA fragments for individual marker were visualized under UV trans-illumination using Molecular Imager^® Gel Doc™ XR System (Bio-Rad Laboratories, Inc., CA, USA).

The genetic structure of the populations was studied using the Bayesian Model-based approach proposed by (Pritchard et al. 2000) using molecular data. The analysis was carried out using the software package, STRUCTURE 2.3. The number of presumed populations (K) was set from 1 to 10, and the analysis was repeated three times. The burn-in period of 100,000 and Markov Chain Monte Carlo replicates of 100,000 and a model with admixture and correlated allele frequencies were used. From the simulation summary of the data run, the best sub-population structure was identified by the posterior likelihood (LnPD) of K which showed a progressive increased until the optimum K, beyond which a plateauing occurs. To pick the optimum K, Structure Harvester (Earl and VonHoldt 2012), an online tool that implemented Evanno’s method of ΔK to pick the threshold posterior probability (Evanno et al. 2005) was used. The run with maximum likelihood was used to assign individual genotypes into groups. Within a group, genotypes with affiliation probabilities (inferred ancestry) ≥ 80% were assigned to a distinct group, and those with < 80% were treated as ‘admixture’, i.e. these genotypes seem to have mixed ancestry from hybrids belonging to same parental lines.

To find out the genetic relationship between different ICR lines, genotypic data were used for the calculation of genetic distance using the Rogers’ distance (Rogers 1972) followed by phylogeny reconstruction using neighbor-joining as implemented in Power Marker v.3.25 (Liu and Muse 2005) with the tree viewed using MEGA X (Kumar et al. 2018). Restorers were identified based on their original group for comparison. The restorer entries were designated with different colour symbols for easy identification in the tree diagram. To estimate the population genetic parameters, the genotypic data were analyzed using the software, PopGene v.1.32 (Yeh et al. 1997) to estimate the F statistics and genetic diversity parameters. The ICR population were grouped based on their origin/the hybrid from which they have been derived. The data were subjected for analysis of molecular variance (AMOVA) using Arlequin 3.1 software (Excoffier et al. 2005). Fixation indices (Weir and Cockerham 1984) and population pairwise F_ST values were computed.

Results

Variation in agro-morphological traits

The analysis of variance indicated the existence of highly significant differences among ICR genotypes for the characters like plant height, number of tillers per plant, days to fifty percent flowering, panicle length, pollen fertility percentage, number of filled grain per panicle, spikelet fertility percentage, test weight and yield per plant (Table 1). All the sources of major variation, restorers, environments and their interaction components showed significant variation. However, the environmental variation was highest for all the traits. Among the traits studied, plant height showed maximum variation (CV of 10.57%), while days to 50% flowering showed the least variation (CV of 0.79%). The mean days to 50% flowering recorded in the restorers was 98 days, which ranged between 89 and 107 days. Plant height on an average ranged between 84.66 and 122.21 cm, while the tillers per plants ranged from 6 to 13 tillers per plant with an average of 10 tillers. Panicle length ranged from 22.6 to 30.2 cm with an average of 26.7 cm, producing about 161.56 fertile grains per panicle. The number of fertile grains varied between 101 and 240 per panicle on average among the restorers, with a coefficient of variation of 3.92%. The mean pollen fertility and spikelet fertility recorded among the restorers was 86.47% and 82.49%, respectively. The weight of 1000 grains varied between 14.57 and 30.66 g, recording an average of 21.42 g in the total population. Grain yield per plant registered a coefficient of variation of 6.77% and an average of 20.91 g, with a range of 14.73–27.62 g per plant (Supplementary Table S1). For all the traits studied, CV was in the low range (0–10%) except for plant height which is in the medium range (10–20%).

Table 1.

Mean squares and summary statistics from combined ANOVA for agro-morphological traits of iso-cytoplasmic restorer lines

Source	Df	DFF	PH	NT	PL	PF	FG	SF	TW	YPP
Environment	2	6915.55*	3089.09*	1064.18*	160.53*	4983.96*	4871.23*	3487.09*	1601.10*	3464.17*
Restorers	99	92.83*	488.53*	13.53*	18.28*	509.17*	5229.39*	144.65*	186.26*	57.90*
Environment × restorer	198	27.02*	245.95*	8.57*	5.21*	142.07*	2837.75*	114.87*	133.23*	54.80*
Average		98.07	102.75	9.85	26.72	86.47	161.56	82.49	21.42	20.91
Minimum		89.10	84.66	6.64	22.60	55.11	101.82	65.61	14.57	14.73
Maximum		106.60	122.21	12.72	30.24	95.43	239.90	91.03	30.66	27.62
Coefficient of variation (CV)		0.79	10.57	8.47	0.87	0.82	3.92	1.58	1.35	6.77
Critical difference (CD)		1.08	15.21	1.17	0.32	0.97	8.81	1.82	14.74	1.98

Df degrees of freedom, DFF days to 50% flowering, PH plant height in cm, NT number of tillers per plant, PL panicle length in cm, PF pollen fertility in %, FG filled grain number per panicle, SF spikelet fertility in %, TW weight of 1000 grains in g, YPP grain yield per plant in g. CV range: 0–10% (low), 10–20% (medium), > 20% (high) (Gomez and Gomez 1984)

*The variances are significant at p < 0.05 level

Relationship among restorer lines based on agro-morphological traits

Based on the average agronomic performance from the combined data, the ICR lines were grouped into two clusters along with the restorer checks (Fig. 1). Among the clusters, cluster 1 was the largest with 62 ICRs, whereas cluster 2 was represented by 38 ICRs. Additionally, cluster 1 contained one of the check restorers, RPHR 1005R, while other two restorer checks, PRR 78 and DRR 714R were found grouped in cluster 2.

Fig. 1.

Neighbour-joining phylogenetic relationship (NJ-tree) among iso-cytoplasmic rice restorers based on agro-morphological traits

Genotypic variations among restorers

The products amplified by the standard SSR markers were in the range of 71–347 bp among the ICR lines (Table 2). Thirty-three markers were polymorphic, while 17 were found to be monomorphic (Supplementary Table S2). Polymorphic SSR markers generated a total of 78 alleles with an average of 2.42 alleles per marker. There were 22 bi-allelic markers (66.7%), eight tri-allelic markers (24.2%) and three tetra-allelic markers (9.10%). RM19, RM171 and RM474 were found to amplify as high as four alleles. Polymorphic markers were predominant on chromosomes 1 and 8 with seven and six markers each, respectively, followed by chromosomes 3, 5 and 7 with three markers each. Chromosomes 6, 9, 11 and 12 were represented by two markers, while chromosomes 2, 4 and 7 had only one marker. The marker statistics showed that the major allele frequency, defined as those alleles with a frequency more than 5%, ranged from 51.0% (RM413) to 99.1% (RM431, RM161 and RM284). Seven markers (RM431, RM161, RM284, RM237, RM259, RM433 and RM44) showed single allele predominance by recording more than 95% of major allele frequency. The observed heterozygosity defined as the percentage of loci heterozygous per individual or the number of individuals heterozygous per locus was observed for 20 out of 32 loci. Highest polymorphism information content (PIC) value of 0.55 was observed for the marker RM 413 followed by RM 474 (0.46) with an average of 0.26.

Table 2.

Group-wise summary of genetic variation statistics based on 33 polymorphic standard GCP SSR markers

Marker*	LG	Min. allele size (bp)	Max. allele size (bp)	AFm	N a	GD	H	PIC
RM1	1	85	115	0.785	2.000	0.308	0.000	0.299
RM495	1	148	160	0.710	2.000	0.412	0.056	0.327
RM259	1	133	186	0.977	2.000	0.046	0.009	0.045
RM312	1	86	106	0.916	2.000	0.154	0.000	0.142
RM5	1	94	138	0.701	3.000	0.425	0.000	0.342
RM237	1	105	153	0.981	2.000	0.037	0.000	0.036
RM431	1	233	261	0.991	2.000	0.019	0.000	0.018
RM154	2	148	230	0.821	3.000	0.300	0.019	0.265
OSR13	3	85	122	0.785	3.000	0.358	0.000	0.327
RM338	3	178	184	0.879	2.000	0.213	0.000	0.191
RM514	3	229	278	0.755	3.000	0.373	0.094	0.307
RM307	4	116	191	0.626	2.000	0.468	0.168	0.359
RM413	5	71	114	0.510	3.000	0.617	0.029	0.547
RM161	5	154	187	0.991	2.000	0.019	0.000	0.018
RM334	5	119	207	0.639	2.000	0.461	0.029	0.355
RM510	6	99	127	0.745	2.000	0.380	0.000	0.308
RM162	6	191	244	0.907	2.000	0.169	0.019	0.155
RM11	7	118	151	0.654	2.000	0.452	0.056	0.350
RM152	8	133	157	0.524	3.000	0.503	0.096	0.381
RM25	8	121	159	0.743	2.000	0.382	0.009	0.309
RM44	8	82	132	0.958	3.000	0.082	0.009	0.080
RM284	8	139	159	0.991	2.000	0.019	0.000	0.018
RM433	8	216	248	0.972	2.000	0.055	0.000	0.053
RM447	8	95	146	0.782	2.000	0.341	0.049	0.283
RM316	9	194	216	0.693	2.000	0.425	0.009	0.335
RM105	9	100	141	0.528	2.000	0.498	0.038	0.374
RM474	10	216	288	0.632	4.000	0.522	0.066	0.459
RM271	10	80	120	0.882	3.000	0.213	0.009	0.199
RM171	10	307	347	0.525	4.000	0.522	0.101	0.410
RM552	11	167	258	0.935	2.000	0.122	0.000	0.115
RM287	11	82	118	0.762	2.000	0.363	0.009	0.297
RM19	12	192	250	0.613	4.000	0.482	0.039	0.375
RM277	12	104	121	0.514	2.000	0.500	0.000	0.375
Mean	–	–	–	0.762	2.485	0.320	0.055	0.264

LG linkage group, AF_m major allele frequency, N_a the average number of allele per locus, GD gene diversity, H observed heterozygosity, PIC polymorphic information content

*Monomorphic markers are presented in the Supplementary Table S3

Population structure

Based on the genetic data from SSR markers, we identified two sub-populations (K = 2) (Fig. 2) with ΔK value of 48.7 (Supplementary Table S3). The sub-populations are named as POP1 and POP2. Populations are divided at the cut off when the affiliation probability (inferred ancestry coefficient) of the lines fell below 0.5 (50%). The lines having affiliation probabilities ≥ 0.8 (≥ 80%) were recognized as less admixed lines and those with probabilities from 0.5 to 0.8 were recognised as admixed lines. POP1 consisted of 52 lines which included all of the seven checks and 45 ICR lines (Table 3). Among these, 15 were admixed, which contained three of the checks, IR 79156B, PRR 78 and IR 58025B. POP2 had 55 lines which included seven admixtures. All the checks fell in POP1 including restorers and maintainers with high affiliation probabilities, except for those showed admixed affiliation. The proportion of admixtures was more in POP1 (28.4%) followed by POP2 (12.7%). The within cluster average distance (expected heterozygosity) was marginally higher in POP2 (0.28), as compared to POP1 (0.27). The mean fixation index (F_ST) was almost similar for both the subpopulations. On the other hand, the proportion of less-admixed lines were more in POP2 (88.3%), followed by POP1 (71.6%).

Fig. 2.

Analysis of the population structure based on 50 SSR markers. An estimated ΔK value for each K value to estimate population structure. The K with the maximum ΔK is taken as the population structure that captures all the variations in the population. Each individual is represented by a vertical bar, partitioned into up to K coloured segments

Table 3.

Population structure of iso-cytoplasmic restorers and checks by Bayesian analysis

Sub-population	Genome constitution	Affiliation probability	Member lines*	F ST	H e
POP1	Less-admixed	0.81–0.99	PRR396, PRR353, PRR394, PRR393, PRR392, PRR390, PRR360, PRR362, PRR398, PRR397, PRR395, DRR714, PRR389, PRR364, PRR370, PRR383, PRR378, PRR368, PRR369, PRR384, PRR399, PRR387, PRR386, PRR350, PRR388, PRR391, PRR349, PRR363, PRR352, PRR382, PRR376, RTN12B, PRR371, PUSA6B, RPHR1005R, PRR354, PRR385	0.267	0.272
	Admixed	0.53–0.80	IR79156B, PRR78, PRR367, PRR321, PRR366, PRR377, PRR365, PRR361, PRR381, PRR379, PRR380, PRR375, PRR373, PRR374, IR58025B
POP2	Less-admixed	0.82–0.99	PRR347, PRR315, PRR345, PRR340, PRR341, PRR306, PRR342, PRR319, PRR318, PRR346, PRR338, PRR344, PRR305, PRR357, PRR308, PRR304, PRR331, PRR325, PRR312, PRR316, PRR337, PRR311, PRR313, PRR307, PRR339, PRR326, PRR343, PRR329, PRR348, PRR333, PRR310, PRR332, PRR356, PRR330, PRR334, PRR327, PRR324, PRR317, PRR351, PRR309, PRR303, PRR302, PRR336, PRR358, PRR301, PRR320, PRR300, PRR323	0.269	0.83
	Admixed	0.52–0.76	PRR359, PRR314, PRR335, PRR322, PRR372, PRR328, PRR355

F_ST mean fixation index, H_e expected heterozygosity

*Checks are indicated in bold

Genetic relations among restorers based on nearest ancestry

Based on the near ancestry that is hybrid from which they were derived, the ICR lines were grouped empirically into two major clusters I and II, comprising of 52 and 55 lines (Fig. 3). Among the checks, IR 79156B, PRR 78, DRR 714R and RPHR 1005R were located in cluster I and RTN 12B, Pusa 6B and IR 58025B were found in cluster II. Cluster I had maximum exclusive representation of ICR lines derived from as many as nine hybrids. ICR lines derived from five hybrids that were found shared between both the clusters, all of them were derived from hybrids released by private seed companies. In cluster II, the entire set of ICR lines obtained from eleven hybrids were found clustered. Although ICR lines derived from the public as well as private sector research organizations were grouped across both the clusters, cluster I had maximum exclusive representation from five public sector hybrids, whereas cluster II had only four. On the contrary, cluster II had total representation from seven private sector hybrids, while cluster I had full set of ICRs derived four public hybrids. No ICRs derived from public sector hybrids were found shared between clusters.

Fig. 3.

Representation of genetic distance among iso-cytoplasmic rice restorer lines based on hybrid origin

Comparing origin-based grouping with population structure

High level of agreement was observed between population structure and the near ancestry-based grouping of ICR lines. The cluster I contained 39 ICR lines that belonged to POP1 (86.7%), and cluster II had 46 ICRs that were in POP2 (83.6%). The admixtures were found grouped in either of the two clusters. When the origins of the restorers were examined (Table 4), POP1 indicated remarkable aggregation of 33 ICRs derived from private sector hybrids, showing 73.3% predominance. The remaining 12 lines (26.7%) were derived from public sector hybrids. In POP2, however, maximum members (36 lines) were derived from public sector hybrids accounting for 65.4% of the sub-population. The remaining 34.5% (19 lines) were derived from private sector hybrids. Eight public hybrids, DRRH 2, DRRH 3, PSD 3, PRH 10, CORH 3, Sahyadri 2, Sahyadri 3 and Sahyadri 4 contributed restorers for POP2 exclusively. Further, all the ICR derivatives of two public hybrids, JRH 8 and KRH 2 were the exclusive members of POP1, while ICRs from private hybrid such as GK 5003, US 312, NK 5251 and PA 6129 were found grouped in POP2. Private hybrids such as PA 6201, PA 6444, PHB 71, Suruchi 5401, JRH 8, JKRH 401, PAC 835 and PAC 837 contributed totally to POP2. Derivatives of Indam 200-017 were found distributed equally across POP1 and POP2 (two lines each) and three ICRs from the hybrid, Sahyadri 1 were found in POP1 and one ICR in POP2. Two hybrids, one public (Indira Sona) and one private (DRH 775) had their three ICRs grouped in POP1 and the remaining one in POP2. All the remaining private-sector hybrid derivatives were found grouped in POP1 and public-sector hybrids in POP2.

Population statistics and analysis of molecular variance

Analysis of molecular variance (AMOVA) revealed that out of the total genetic variance among ICR lines, 40.88% was attributed to the groups based on their origin (hybrids from which they were derived), and the remaining 59.12% was explained by variation among individuals within populations (50.38%) and within individuals (8.74%) (Table 5). Correspondence between group variance was 1.99, within group variance was 2.45 and within individual (residual) variance was 0.43.

Table 5.

Analysis of molecular variance (AMOVA) of iso-cytoplasmic restorer lines of rice

Source of variation	Df	Sum of squares	Variance components	Percentage of variation
Among populations	26	548.12	1.99	40.88
Among individuals within populations	80	426.08	2.45	50.38
Within individuals	107	45.50	0.43	8.74
Total	213	1019.71	4.86

Df degrees of freedom

The statistics of the derived populations-based hybrid origin indicated significant differentiation among all the pairs of sub-groups with pairwise F_ST values ranging from 0.10 to 0.81 with an average value of 0.44 (Fig. 4). Some sets showed significant differentiation from rest of the population, such as lines derived from DRRH 2, Sahyadri 2, DRH 775, PA 6129, Indira Sona, PAC 837 and CORH 3, showed high differentiation with respect to each other or with rest of the population.

Fig. 4.

Pairwise F_ST matrix of iso-cytoplasmic restorer lines and checks

Allele distinctiveness among the restorer groups based on their pedigree

From the allelic profile of 32 polymorphic SSR markers, distinct allelic patterns could be recognized for the ICR lines derived from certain hybrids in the study (Table 6). Restorer derivatives of DRRH2, PSD 3, Indam 200-017, PA 6201, PA 6444, PHB 71, Indira Sona and KRH 2 were distinguished by two markers each, while the hybrid derivatives from Sahyadri 3, DRH 775, JRH 8 and JKRH 401 were distinguishable by one marker each.

Table 6.

Distinct marker patterns unique to restorer populations derived from specific hybrids

Hybrid	Distinct marker alleles
DRRH 2	RM152 (170 bp), RM152 (140 bp), RM514 (280 bp)
PSD 3	RM287 (120 bp), RM514 (280 bp)
Sahyadri 3	RM144a (250 bp)
Indam 200-017	RM1 (120 bp), RM413 (100 bp)
DRH 775	RM152 (130 bp)
PA 6201	RM514 (260 bp), RM514 (240 bp), RM316 (210 bp)
PA 6444	RM413 (75 bp), RM152 (150 bp)
PHB 71	RM19 (220 bp), RM263 (210 bp), RM263 (180 bp)
Indira Sona	RM495 (150 bp), RM316 (210 bp)
JRH 8	RM237 (130 bp)
JKRH 401	RM152 (150 bp)
KRH 2	RM514 (260 bp), RM316 (210 bp)

Discussion

Diversification of restorers is one of the major activities in hybrid rice breeding, because of the limitations in diversifying male sterile lines. WA-CMS in rice is expressed only when the WA cytoplasm interacts with the nuclear genes that favour maintenance of pollen sterility, which restricts the genetic variability among the CMS lines. However, for the exploitation of maximum level of heterosis or hybrid vigour, it is essential that the parents be genetically diverse (Bar-Hen et al. 1995; Rajendran et al. 2012). Using diverse restorers is one of the alternatives to address this limitation in hybrid breeding. Although there is ample variability for restorer identification in the rice gene pool, it is essential that restorers for WA-CMS are to be sourced within indica sub-group, as the Rf genes are absent in japonica types (Ikehashi 1982). Further, the presence of restorer genes in a line alone may not be useful in hybrid development, because the heterotic potential of hybrids results from the interaction at different genomic loci. Grouping or classification of genotypes based on fertility restoration ability and heterotic potential (better combining ability) is the prime objective of development of heterotic pool in rice (Beukert et al. 2017). Restorers can be developed from the existing gene pool by screening the available germplasm for fertility restoration ability or by selection from the segregating populations involving popular hybrids. ICR development from popular hybrids is advantageous, since they possess the same cytoplasm of the hybrid (iso-cytoplasmic), from they are derived, and therefore, the full complement of Rf genes for the restoration of fertility. Since WA-CMS is based on the most widely used WA cytoplasm in hybrid breeding, the restorers developed from WA-CMS hybrids can effectively nullify any interaction effect between WA cytoplasm and nuclear genes (Kumar et al. 2017b). Moreover, it offers a possibility to recover a good assembly of heterotic loci as they are derived from already popular high yielding hybrids with high heterosis. These ICR lines with high per se and stable performance across environment can be directly utilized in suitable scale developing heterotic hybrids.

The 25 commercial hybrids from which ICRs in the current study were derived were reported to be genetically diverse (Anand et al. 2012; Singh et al. 2016). All these hybrids were evaluated by Anand et al. (2012) and reported that they fell into two clusters, wherein 16 hybrids were grouped into one cluster and eight in the second and one remained ungrouped. This study reported close genetic resemblance of private sector hybrids such as JKRH 401 and PA 6444, US 312 and GK 5003, and among public sector hybrids, JRH 8 and DRRH 2. Besides, nine private sector hybrids were found grouped with seven public sector hybrids in cluster 1, whereas four each of public and private sector hybrids shared cluster 2. In the second study, Singh et al. (2016) used 14 of the hybrids used in this study and reported that private and public-sector hybrids shared two clusters in mixed combinations. Although the information on the parentage of private hybrids is available, all of them are coded and therefore obscure for the comparison with parental information of public sector hybrids. Moreover, in case of public sector hybrids, the restorer parents are often code named to hide identity, although the CMS lines are generally disclosed. Because of this, previous studies have concluded that most of the parents of the public-sector hybrids share limited genetic diversity, especially on the maternal lineage (as there are limited WA-CMS sources), and this must also hold good for private-sector hybrids.

The ICR lines developed in this study displayed morphological variation, especially for plant height and grain yield, which will be advantageous for hybrid rice improvement. Plant height is an important trait in hybrid rice development, because restorers which are relatively taller than the CMS lines offers the advantage of efficient pollination in hybrid seed production, thereby improving hybrid seed yields. Further, plant height is also important in yield determination in rice (Sabouri et al. 2008), with semi-tall plants yielding better than dwarf types. Similarly, grain yield also showed apparent variation among the ICR lines indicating that they are harbouring variable gene combinations that can throw differential yield expression. This is advantageous in hybrid development as the variable combination can potentially throw greater heterosis for grain yield. The variation is moderate for other traits, which could be due to selection pressure exercised on the ICRs for plant yield and its component and panicle exsertion. Further, grouping based on morphological traits has grouped the restorer lines into two major groups, which is expected, because all the lines are derivatives of elite hybrids. Moreover, origin-based grouping with respect to morphology is unexpected as all the hybrids used are semi-dwarf to semi-tall high yielding ones, with lesser apparent morphologic variations. Moreover, recombination events occurred, while stabilizing selection of ICR lines, would have shuffled genes for various morphological traits, leading to a random pattern of gene assembly, which could diminish the variation between the lines. It has been established that genetic divergence analysis among rice genotypes on the basis of morphological traits can be used to classify and differentiate different genotypes in a population (Franco et al. 2001) provided large contrast existed between several traits. If such contrasts are limited to few traits such as observed in this study, the effectiveness of quantitative or morphological trait-based grouping will be limited to fewer groups. Although genetic divergence analysis plays a vital role in the selection of diverse genotypes (Shahidullah et al. 2010), it is important to use most contrasting and independent traits to get better dispersion and cluster resolution, so that the selection of members between the clusters would ensure a divergent genotype assembly and broad spectrum of variability (Sarawgi and Rastogi 2000; Roy et al. 2002; Naik et al. 2004).

The assessment of genetic variation by use of expressed traits has limitations of unaccounted variations. These variations are manifested by a pattern of gene expression, gene–environment interactions, gene–gene interactions, besides external factor such as environmental effects. Often partitioning of such variations becomes difficult in practice, which jeopardizes the efficiency of using trait-based variation for genetic diversity assessment (Murren et al. 2015). On the contrary, molecular genetic variations are unaffected by such factors both internal and external and are recognized as the best tools for assessing genetic variation among the genotypes. However, total genotypic variations comprise of both expressed and unexpressed sequence variations as well as variations at untargeted loci that may hoodwink the variations with respect to a specific trait (Zhang et al. 2010). Despite this, the total genetic variations estimated by a random set of markers have been found as useful resources for genetic diversity estimation and does not deviate much from the diversity estimated by the trait linked markers (Yadav et al. 2013).

The overall genetic divergence of ICR lines revealed that the genetic variability among these lines was moderate, which could be ascribed to the genetic closeness of parental combinations used in developing these hybrids (He et al. 2012) and also on the selection exercised for grain yield. In the current study, the average number of alleles (2.42 alleles/locus) is lesser than the average number of alleles (3.88 alleles/locus) reported by Zhang et al. (2011) in rice core collection with 150 rice varieties from South Asia and Brazil, but quite comparable to values reported for studies performed on smaller germplasm sets (Hashimoto et al. 2004; Siwach et al. 2004). However, since the ICR lines are mainly from the genetic background of adapted popular rice hybrids, the allelic diversity in these lines can be considered to be substantial enough to enable the development of improved rice hybrids. The higher value of gene diversity among the ICR lines at certain loci could serve as signatures of reproductive isolation of parental stocks (Young et al. 1996; Baack et al. 2015) that could have occurred in several ways in the vast rice gene pool. Further, rice is a self-pollinated crop and intermating is seldom possible among different plants that are geographically isolated (Breseghello and Coelho 2013). Therefore, ICRs can serve as a useful supplementary approach for restorer diversification, which recycles the already utilised genetic diversity and heterotic potential into newer recombinants.

Model-based approach to identify the underlying sub-population structure based on the assumptions of shared co-ancestry of admixed populations is frequently implemented various researchers in population genetic studies (Jin et al. 2010; Zhang et al. 2011; Courtois et al. 2012; Reig-Valiente et al. 2016). Although ICR lines in the present study were derived from 25 hybrids, genotypic data could classify them only into two sub-populations. This validated the basic assumption of the model analysis that the hybrids were sharing common co-ancestry. However, the individual populations have achieved considerable uniqueness with respect to each other, which was also indicated by the lower average value of heterozygosity (0.055), which gives an indication that lines have achieved substantial homozygosity after six generations of selfing. Eighty-five ICR lines were having affiliation probabilities ≥ 80% and thus were genetically distinct with respect to the populations they represented. It may also be emphasized that the lines showed little admixture between populations. Ancestry threshold of 80% was also used to identify accessions belonging to a specific subpopulation (Courtois et al. 2012).

15% of the lines had admixed constitution, with POP1 having the maximum admixing (12 lines) than the other population, implying that POP2 had more overlapping genetic admixing with POP1. POP1 formed a relatively small group of admixed lines (seven lines) with a major contribution of ICR lines from two public hybrids, KRH 2 and JRH 8. Rest of the public hybrids were found associated to POP2. Earlier report on genetic closeness of hybrids indicates that Pusa Rice Hybrid 10 and Sahyadri 1 are clustered in one group, which DRRH 2 was located in another group (Anand et al. 2012). It is known that the female parent of Pusa Rice Hybrid 10, Pusa 6A shares the lineage derived from IR 58025A, the female parent of Sahyadri 1. Pusa 6A is developed by crossing with IR 58025A to transfer the WA cytoplasm to a perfect maintainer background of Pusa 150-21-1-1 (Nagarajan et al. 2012). IR 58025A is an aromatic line developed by International Rice Research Institute (IRRI) from the line, Pusa 167-120-3-2 by crossing it with IR48483A to transfer WA cytoplasm (Singh 2000). This line was extensively used in hybrid development in Asian countries (Xie 2009) especially in India, because of its excellent combining ability. Because of the extreme popularity of IR 58025A, several first-line Indian rice hybrids share common maternal lineage from this line. In this study, it is known that ICR lines developed from the different hybrids namely KRH 2, Sahyadri 1, Sahyadri 2, Sahyadri 3, Sahyadri 4, Indira Sona share the maternal lineage of IR 58025A. However, KRH 2 was found associated to POP1 and Indira Sona was found associated to both the populations. This indicates the limitation of maternal diversity in classifying the ICRs. Besides, there could be other hybrids from the public sector, sharing similar lineage but remains obscured due to limited parental information (Anand et al. 2012). Considering the limitations in maternal diversity in the present set of ICR lines, the divergence and sub-population isolation observed between the sub-populations could be attributed predominantly to the restorer parents. The exclusive congregation of ICR lines from private-sector hybrids in POP1 and public-sector hybrids in POP2, could also be possible due to their linked ancestry. Therefore, as observed in the case of the maternal lineage of hybrids, there may also common paternal lineage among the hybrids as suggested in the case of hybrids, US 312 and GK 5003 (Anand et al. 2012).

Genetic differentiation among ICR lines derived from hybrids that shared common maternal parent, IR 58025A throws light on the paternal contribution of genetic diversity. The restorers developed from Sahyadri 1, Sahyadri 2, Sahyadri 3 and Sahyadri 4 were found grouped in POP2, while POP1 contained restorers from Indira Sona and KRH 2, of which ICR lines from Indira Sona were found distributed between both the sub-populations. This is possible only due to genetic diversity from the paternal lineage of these hybrids, which is an indirect indication that the restorer parents of the hybrids used in this study could have been derived from diverse sources. The sustained success of the hybrid breeding program lies on identifying the heterotic groups in the parental gene pool (Melchinger and Gumber 1998). DNA fingerprinting and genetic diversity among hybrid rice parental lines using simple sequence repeat (SSR) markers have been considered robust, due to their characteristics such as being highly polymorphic, multi-allelic and co-dominant (He et al. 2012).

POP2 was more genetically isolated from POP1, based on the presence of admixtures. Therefore, restorer parents of most public-sector hybrids would be more genetically distinct than the rest of the hybrids. Therefore, the ICRs from these distinct sub-populations, followed by POP1 members could be more genetically diverse towards restorer diversification, which fulfils the prime aim of this study for development and identification of diverse restorers for development of future heterotic hybrid combinations.

Grouping based on the nearest ancestry of the ICR lines was well in agreement with the sub-population differentiation. The population genetic parameters of this grouping have identified certain hybrid groups which are more differentiated than others. The hybrids such as DRRH 3, PSD 3, PRH 10, CORH 3, Sahyadri 3, GK 5003, US 312 and NK 5251 formed a genetically more closer set in POP2, which was diverse from the group in POP1 containing hybrids such as, PHB 71, PAC 837 and KRH 2. The parental combinations of those hybrids that are more genetically distant than the rest of the hybrids can throw more diverse ICRs. The ICR lines developed from these hybrids are genetically distinct, which will be utilized in future hybrid development in rice.

Conclusion

The ICR lines evaluated in the present study have been found to harbour moderate genetic variation and distinct divergence, indicating their potential for direct utilization as restorers in hybrid rice breeding involving WA-CMS system. These ICRs were found to be grouped into three subpopulations with good population differentiation, possibly gained from their respective paternal lineage. Although they have enough genetic properties for use as restorers in the hybrid rice breeding, there is a need to characterize their combining ability and heterotic potential for further use. Test of fertility restoration ability and hybrid potential of these lines in a separate study indicates improved fertility restoration ability for ICR derived hybrids (Kumar et al. unpublished data). The success in developing an array of diverse restorers in the present study can be broadened by including more number of recently released hybrids, which can help in strengthening the ongoing efforts on restorer diversification in rice breeding. Here, we prove that this approach is workable in identifying effective restorers that are recombinants of diverse genetic backgrounds.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

Conceptualization of research (AKS, GKS); Designing of the experiments (AKS, GKS, AK); Contribution of experimental materials (AK, PKB); Execution of field/lab experiments and data collection (AK, GKS, VJS, PKB, MN); Analysis of data and interpretation (AK, GKS, KKV, PKB, AKS); Preparation of manuscript (AK, GKS, KKV, PKB, AKS, RKE, HB).

Compliance with ethical standards

Conflict of interest

The authors declare no conflict of interest.