Skip to main content

Genomic introgressions from African rice (Oryza glaberrima) in Asian rice (O. sativa) lead to the identification of key QTLs for panicle architecture



Developing high yielding varieties is a major challenge for breeders tackling the challenges of climate change in agriculture. The panicle (inflorescence) architecture of rice is one of the key components of yield potential and displays high inter- and intra-specific variability. The genus Oryza features two different crop species: Asian rice (Oryza sativa L.) and the African rice (O. glaberrima Steud.). One of the main morphological differences between the two independently domesticated species is the structure (or complexity) of the panicle, with O. sativa displaying a highly branched panicle, which in turn produces a larger number of grains than that of O. glaberrima. The gene regulatory network that governs intra- and interspecific panicle diversity is still under-studied.


To identify genetic factors linked to panicle architecture diversity in the two species, we used a set of 60 Chromosome Segment Substitution Lines (CSSLs) issued from third generation backcross (BC3DH) and carrying genomic segments from O. glaberrima cv. MG12 in the genetic background of O. sativa Tropical Japonica cv. Caiapó. Phenotypic data were collected for rachis and primary branch length, primary, secondary and tertiary branch number and spikelet number. A total of 15 QTLs were localized on chromosomes 1, 2, 3, 7, 11 and 12, QTLs associated with enhanced secondary and tertiary branch numbers were detected in two CSSLs. Furthermore, BC4F3:5 lines carrying different combinations of substituted segments were produced to decipher the effects of the identified QTL regions on variations in panicle architecture. A detailed analysis of phenotypes versus genotypes was carried out between the two parental genomes within these regions in order to understand how O. glaberrima introgression events may lead to alterations in panicle traits.


Our analysis led to the detection of genomic variations between O. sativa cv. Caiapó and O. glaberrima cv. MG12 in regions associated with enhanced panicle traits in specific CSSLs. These regions contain a number of key genes that regulate panicle development in O. sativa and their interspecific genomic variations may explain the phenotypic effects observed.

Peer Review reports


Improving or enhancing the sustainability of rice yield continues to be a crucial challenge in the breeding of this crop, especially in the context of a growing world population and with regard to climate change. Inflorescence architecture in rice directly affects yield potential through the regulation of grain number. Grain number per panicle depends on orders and numbers of branches (i.e., primary, secondary and potentially tertiary branches as well as both lateral and terminal spikelets) and on the length of each axis. Rice panicle architecture is based on the production of a series of lateral meristems with distinct identities [1]. After the floral transition, the Shoot Apical Meristem (SAM) is converted into an indeterminate Rachis Meristem (RM) which produces several axillary meristems, the primary branch meristems (PBMs) that subsequently give rise to primary branches (PBs). Once the RM has lost its activity, the newly formed PBM elongates and initiates a variable number of axillary meristems. These acquire the identity of secondary branch meristems (SBMs) and develop into secondary branches (SBs), which in turn may produce tertiary branch meristems (TBMs), or alternatively differentiate directly into lateral spikelet meristem (SpMs) and then florets. The basic architecture of the panicle is thus determined by patterns of axillary meristem formation and the specification of their identities. Overall, the complexity and diversity of panicle branching can be considered as governed by two key elements: the number of axillary meristems produced during the indeterminate phase; and the rate of meristem fate transition, which determines whether an axillary meristem grows into a higher-order branch or differentiates into a spikelet.

Over recent decades, a number of genes and quantitative trait loci (QTLs) associated with panicle development and affecting its architecture have been identified and functionally described in rice [2,3,4,5,6,7,8]. Regulatory genes influencing grains number per panicle include Gn1a/OsCKX2 (GA20-oxidase1, LOC_Os03g63970 [9]), IPA1/WFP (OsSPL14 SQUAMOSA PROMOTER BINDING PROTEIN-LIKE, LOC_Os08g39890 [10, 11]), LAX1 (basic-helix-loop-helix, LOC_Os01g61480 [12]), PAP2 (MADS box, LOC_Os03g54170 [13, 14]), DEP1 (G protein gamma subunit, LOC_Os09g26999, [15]), MOC1 (GRAS family nuclear protein, LOC_Os06g40780 [16]), TAW1 (ALOG, LOC_Os10g33780 [17]), qSrn7/FZP (AP2, LOC_Os07g47330 [18,19,20,21]) and APO1 (F-box protein, LOC_Os06g45460 [22]), which positively regulate the numbers of primary and secondary branches. Many of these genes and QTLs encode transcriptions factors, such as those belonging to the SQUAMOSA PROMOTER BINDING PROTEIN-LIKE (SPL), APETALA2/ETYLENE RESPONSE FACTOR (AP2/ERF), MADS-BOX, HOMEOBOX and ALOG families, and are implicated in axillary meristem formation or in the conversion of indeterminate meristems to spikelets, thus influencing the branching complexity of the panicle [12, 13, 20, 23, 24]. For example, the SQUAMOSA PROMOTER BINDING PROTEIN-LIKE OsPL14 (also known as IDEAL PLANT ARCHITECTURE1, IPA1 and WHEALTHY FARMERS's PANICLE, WFP) promotes panicle branching [10, 11], while FRIZZY PANICLE (FZP also known as qSNR7), an ethylene-responsive element binding factors, promotes SM identity [20, 25]. Variations in their regulation lead to significant effects on panicle architecture [10, 19, 26].

Most studies of rice panicle development have focused on the Asian crop plant Oryza sativa L. However, a second species of cultivated rice, O. glaberrima Steud, was domesticated independently of Asian rice, in the inner delta of Niger river [27] from the wild relative species O. barthii around 2,500 years ago. Cultivation of O. sativa was introduced into West Africa around 400 years ago and this crop has since largely replaced O. glaberrima – although the latter is still grown, it represents only 1–2% of the total cultivated area. O. glaberrima is more adapted to a variety of ecologically and climatically diverse regions, its agronomically useful characters including higher tolerance to drought, high temperatures and infertile soils as well as a greater resistance to various biotic stresses [28,29,30,31,32,33]. While O. sativa is less adapted to the African environment, this species has a higher yield potential than O. glaberrima, the difference being partly explained by the higher panicle complexity of Asian rice [34]. Even if molecular pathways associated with the regulation of axillary meristem identity seem to be conserved between the two species, it was possible to identify a set of genes displaying differential expression in relation to panicle phenotypic variability between African and Asian rice [34]. Moreover, a Genome-Wide Association Study (GWAS) carried out on a panel of African rice accessions allowed the identification of several new genomic regions associated with panicle branching diversity and climatic variables [35].

In the context of climate change, achieving high tolerance to environmental stresses is of paramount importance to rice agriculture while knowledge relating to the genetic control of traits governing yield potential in African rice remains very limited. Agronomic traits such as heading date, yield, plant height and grain size are controlled by many genes exerting major and/or minor effects. Identifying the genes and genomic regions associated with interspecific variation in the number of grains produced per panicle is challenging due to the polygenic nature and environmental sensitivity of panicle branching traits.

Libraries of introgression lines (ILs) provide a useful means to investigate genetic phenomena in an interspecific context. An ILs library is a collection of lines with a common genetic background (i.e. the recipient genome), each carrying one or a few genomic regions originating from a donor genome. The ILs are generally obtained through recurrent backcrossing onto the recipient parent, followed by several rounds of self-fertilization (BCnFn) or double haploidization (BCnDH). The genome of the donor genotype is then represented in the library by discrete homozygous chromosomal fragments. Such ILs are also called Chromosome Segment Substitution Lines (CSSLs). In contrast to segregating populations, such as Recombinant Inbred Lines (RILs), Doubled Haploid (DH), Backcross (BC) or F2:3 populations, QTL analyses using CSSLs are suitable to evaluate minor allelic differences conferred by additive QTLs in a uniform genetic background in populations of small sizes [36, 37]. Genetic dissection of complex traits can thus be achieved by combining genetic variation with introgressed genomic fragments so as to reduce interference effects between QTLs. Hence, CSSLs provide a powerful means to identify QTLs for complex traits with minor and/or additive effects [38, 39]. This approach may provide access to novel and potentially beneficial genes "hidden" in the genetic background of a related species that can be discovered when placed in the genetic background of a cultivated species.

In this study, we evaluated several panicle morphological traits in a BC3DH CSSL population developed from an interspecific cross between the recipient parent O. sativa ssp. japonica (cv. Caiapó) and the donor parent O. glaberrima (cv. MG12) [40]. QTLs relating to modifications in panicle morphology were mapped on the rice genome, and 4th generation backcross (BC4F3:5) generations were produced to evaluate the effect of each region on panicle phenotype variation. Finally, DNA polymorphisms between the two parental genomes were investigated in detail for some key genes that were previous implicated in panicle branching diversity in O. sativa. The results obtained will help allow the identification of functionally significant polymorphisms and, more generally, provide an insight into the genetic bases of panicle architecture diversity between the two species.


Panicle traits in the CSSL population

In order to identify new genetic factors governing the panicle branching diversity observed between O. sativa cv. Caiapó (hereinafter referred to as Os_Caiapó) and O. glaberrima cv. MG12 (hereinafter referred to as Og_MG12), phenotypic measurements were performed on a population of 60 BC3DH CSSLs and their parents (Additional file 1: Table S2) [40].

Six quantitative traits were evaluated per panicle: rachis length (RL); primary, secondary and tertiary branch numbers (PBN, SBN, TBN respectively); primary branch length average (PBL); and spikelet number (SpN) (Table 1, Additional file 1: Table S2). The two parents showed contrasting panicle phenotypes with a higher panicle complexity observed in Os_Caiapó compared to Og_MG12. More specifically, the Os_Caiapó parent displayed panicles with more PBN and SBN leading to a higher SpN compared to Og_MG12 (Fig. 1a). In contrast, the Og_MG12 parent produced panicles with longer PBs compared to Os_Caiapó. Since the trials were conducted as two repeated experiments, we computed broad sense heritability for each experiment (Table 1). All the measured variables showed broad sense heritability values higher than 0.8 except for TBN. For all traits, the mean of the CSSL population was similar to the mean of the recurrent Os_Caiapó parent (Table 1). The coefficient of variation for the SBN/PBN ratio was higher in the CSSL population than in Os_Caiapó (26.13% versus 18.26%). Histograms of the CSSLs for panicle traits showed a continuous distribution for all traits except TBN (Fig. 1b). The observed distributions were similar between the two repeated experiments. The abnormal distribution of the TBN trait was associated with a high value of CV (Coefficient Variation) and a low heritability value. This observation results from the rare and unstable nature of this trait which is dependent on environmental conditions and does not appear frequently in the Os_Caiapó parent. Moreover, the formation of tertiary branches has not been described previously in O. glaberrima populations [35], nor was it observed in our earlier experiments.

Table 1 Variation, descriptive statistics and broad sense heritability of the scored traits in the CSSL population and the parents
Fig. 1
figure 1

Panicle structure and morphological panicle trait values in the two parental lines and the 60 BC3DH CSSLs. a Contrasted spread panicle between the two parental lines (Os_Caiapó and Og_MG12). Panicle traits measured are showing in the image: the rachis is the main and central axis of the panicle, rachis length (RL) is measured between the 2 dot black points. The primary branches (PB) (in blue) are axis attached to the rachis, the primary branches bear secondary branches (SB) (in red) which bear tertiary branches (TB) (in yellow). Spikelets (Sp) are attached to the branches. The Primary branch Length (PBL) is the average of the primary branch lengths in the panicle. b Distribution of panicle trait values in the 60 BC3DH CSSLs. Dashed vertical lines correspond to the mean values of each parent (green for Og_MG12 and yellow for Os_Caiapó). Values for repetitions 1 and 2 are shown in light gray and dark gray, respectively. Abbreviations: RL, Rachis Length; PBN, Primary Branch Number; PBL, Primary Branch Length; SBN, Secondary Branch Number; TBN, Tertiary Branch Number; SpN, Spikelet Number

By analyzing relationships between the different panicle traits, we observed the highest correlation (0.71) between SpN and SBN/PBN (Fig. 2a). Other traits were moderately correlated with each other, notably PBN and SpN. We also observed a low correlation between PBL and SBN. Principal component analysis showed a separate distribution of Os_Caiapó and Og_MG12 parents on the PC1 and PC2 axes. In contrast, we observed an overlapping of the BC3DH lines with the Os_Caiapó recurrent parent (Fig. 2b). Analysis of variable contributions showed that SpN is the main trait contributing to the diversity observed in this population.

Fig. 2
figure 2

Description of the morphological panicle traits observed in the BC3DH CSSL lines and parents. a Correlation plot of panicle morphological traits based on the Pearson method between panicle traits, ** represents the p-values of Pearson correlation coefficients < 0.01; *** < 0.001. b Principal component analysis with trait contributions and individual panicle distributions. Abbreviations: RL, Rachis Length; PBN, Primary Branch Number; PBL, Primary Branch Length; SBN, Secondary Branch Number; TBN, Tertiary Branch Number; SpN, Spikelet Number

Detection of QTLs associated with panicle architecture traits

A Dunnett's test revealed a total of 37 lines with significant differences compared to the Os_Caiapó recurrent parent for all traits combined (Additional file 1: Table S4). In general, these lines bear at least two Og_MG12 introgression fragments. For this reason, assigning the underlying QTLs to a single chromosome segment was not straightforward. The CSSL Finder software was therefore used to detect genomic regions associated with panicle phenotype variation with an F-test at each marker [40]. Graphical genotype representations were also produced in which the CSSLs were ordered by trait value for each evaluated trait (Additional file 2: Figure S1). We considered the F-test as significant when its value was higher than 10.0 (\(p-value<\sim 0.002 based on Bonferron{i}{\prime}s test correlation 0.05/200\)). QTLs were assigned to the Og_MG12 introgressed regions of these CSSLs if the F-test was significant and confirmed by the graphical genotype logical analysis. In total, 15 QTLs were detected for all combined traits analyzed with the exception of rachis length (Table 2, Additional file 2: Fig. S1, S2). The F-test value of the markers allowing the detection of each QTL are reported in Table 2 along with their positions in the O. sativa and O. glaberrima reference genomes (cv. Nipponbare IRGSP-1.0 and cv. CG14 OglaRS2 respectively).

Table 2 QTLs for morphological panicle traits detected in the CSSL population

RL trait: 13 CSSLs showed significant changes in RL value in comparison to the Os_Caiapó recurrent parent (Additional file 1: Table S4). However, no QTLs were detected for this trait (Additional file 2: Figure S1a).

SBN/PBN trait: a total of 16 CSSLs showed a significant difference in secondary branch number per primary branch (SBN/PBN ratio) compared to the Og_MG12 parent (Table S4). Two QTLs associated with a decreased SBN/PBN ratio were detected (qSBN/PBN11 and qSBN/PBN12, maximum F-test scores of 15.09 and of 14.94 respectively) (Table 2; Additional file 2: Figure S1b). The average SBN/PBN ratio values were 3.0 for Os_Caiapó and of 1.7 for Og_MG12, corresponding to a decrease of 42% in Og_MG12 relative to Os_Caiapó. The CSSLs L_42, L_55 and L_56 showed a decrease of 44.7%, 42.4% and 36.3% respectively relative to the Os_Caiapó parent, suggesting a high effect of the Og_MG12 introgression(s) in these lines (Additional file 1: Table S4). The phenotyping of these lines was repeated for a second year to confirm panicle trait variation (Additional file 2: Figure S3). The L_55 and L_56 lines contain an Og_MG12 segment in chromosome 11, but with missing marker information at the position of qSBN/PBN12. In contrast, the CSSL L_42 displays only an Og_MG12 segment in chromosome 12 (Additional file 2: Figure S3).

SpN trait: a total of 12 CSSLs showed a significant difference compared to the Os_Caiapó recurrent parent (Additional file 1: Table S3). Two QTLs (qSpN3 and qSpN11) were associated with a decreased SpN with maximum F-test scores of 10.8 and 15.1 respectively (Table 2; Additional file 2: Figure S1c). CSSLs L_55, L_11 and L_26 showed decreases of 37.9%, 32.3% and 30.8% respectively compared to the Os_Caiapó parent (Additional file 1: Table S4). Lines L_55 and L_26 contain Og segments in both chromosomes 3 and 11 in the positions of the detected QTLs (Additional file 2: Figure S3). Line L_11 contains only one Og_MG12 introgression in chromosome 11, related to qSpN11 (Additional file 2: Figure S3).

PBN trait: a total of 13 CSSLs had significantly different PBN values compared to the Os_Caiapó recurrent parent (Additional file 1: Table S4). Three QTLs corresponding to a decreased PBN (qPBN1, qPBN2 and qPBN7) with maximum F-test scores of 12.56, 15.49 and 16.44 were detected on chromosomes 1, 2 and 7 respectively (Table 2; Additional file 2: Figure S1d). Average PBN values were 11.9 for Os_Caiapó and 8.2 for Og_MG12 (Additional file 1: Table S4). The PBN values of two CSSLs (L_10 and L_46) differed significantly from the Os_Caiapó recurrent parent with a reduction of 24.1% and 20.9% respectively (Additional file 1: Table S4). These two CSSLs harbor similar Og_MG12 segments in chromosomes 1, 2 and 7 (Additional file 2: Figure S3).

PBL trait: Fourteen CSSLs exhibited significant differences with the Os_Caiapó recurrent parent for the PBL trait. Three QTLs corresponding to an increased PBL (qPBL1, qPBL3 and qPBL7) with maximum F-test scores of 22.55, 14.51 and 15.78 were detected on chromosomes 1, 3 and 7 respectively (Table 2; Additional file 2: Figure S1e). Average PBL values were 11.1 for Os_Caiapó and 12.3 for Og_MG12 (Table 1). Two CSSLs (L_10 and L_46) showed increased PBL values of 19% and 12.8% respectively in comparison to the Os_Caiapó recurrent parent and harbored similar Og_MG12 segments in chromosomes 1, 3 and 7 (Additional file 1: Table S4; Additional file 2: Figure S3).

TBN trait: Surprisingly, several CSSLs showed a significantly increased TBN value in our field conditions (Additional file 1: Table S3) and five QTLs were detected: qTBN1, qTBN3-1, qTBN3-2, qTBN5 and qTBN7 with maximum F-test scores of 31.11, 21.91, 15.98, 25.12 and 66.09 respectively (Table 2; Additional file 2: Figure S1f). The L_10 and L_4 lines have similar Og_MG12 segments in chromosomes 1, 3 and 7. Within chromosome 5, only the L_4 line contains a segment from the Og_MG12 parent (Additional file 2: Figure S3). As the L_46 line contains otherwise similar Og_MG12 introgressions in chromosome 1, 3 and 7, this CSSL was included in the second phenotyping campaign, in which it revealed the presence of tertiary branches on its panicles in contrast to L_4 (Additional file 2: Figure S3). Overall, these results support an association of Og_MG12 introgressions with the presence of tertiary branches.

For the PBN, PBL and TBN panicle traits, the same BC3DH CSSLs (L_10 and L_46) led to the detection of several QTLs. The two lines in question display a similar panicle phenotype with the presence of tertiary branches associated with a decreased PBN and increased PBL and SBN/PBN ratio values compared to the Os_Caiapó recurrent parent (Fig. 3a). The L_10 and L_46 BC3DH lines contain a complex association of Og_MG12 introgressions in their genomes (Fig. 3b). For clarification, the regions corresponding to colocalized QTLs, associated with different traits, were renamed without specifying the associated traits: q_1 (for qPBL1, qPBN1 and qTBN1); q_2 (for qPBN2), q_3-1 (for qTBN3-1); q_3-2 (for qPBL3 and qTBN3-2); q_5 (for qTBN5); and q_7 (for qTBN7, qPBN7 and qPBL7).

Fig. 3
figure 3

Phenotypic and genetic description of BC3DH and BC4 lines to dissect the effects of QTLs on panicle traits. a Schematic representation of the panicle structure observed in Os_Caiapó, Og_MG12 and BC3DH L_10 and L_46, indicating primary branch (PB) in black, secondary branch (SB) in yellow, tertiary branch (TB) in red and spikelet (Sp) in green. Scale bar = 1cm. b Boxplots of the phenotypic variation observed in Os_Caiapó (yellow), Og_MG12 (green), BC3HD (red) and BC4 (blue) lines. Each point represents the phenotypic value for one panicle. Statistical significance (t-test p-values) between Os_Caiapó and each line for the panicle morphological traits is indicated as follows: ** p-values < 0.01; *** < 0.001. The left-hand panel represents the allelic status (Os_Caiapó in yellow or Og_MG12 in green) for each QTL region in each line. c Visualization of Og_MG12 segments present in chromosomes 1, 3, 5 and 7 in BC3DH and BC4 lines to dissect the effects of QTLs on panicle traits. The positions of QTLs associated with PBN, PBL and TBN traits are indicated immediately above them (bars shaded in orange, red and blue respectively). Abbreviations: PBN, primary branch number; PBL, primary branch length; TBN, tertiary branch number; SBN/PBN, ratio between secondary branch and primary branch numbers; SpN, spikelet number

Substitutions of Os_Caiapó genomic regions by corresponding Og_MG12 segments can result in an added value panicle phenotype

As the altered phenotype observed in lines L_10 and L_46 was associated with multiple substitution segments (Fig. 3b), it was critical to determine whether this phenotype, and notably the formation of tertiary branches, involved one or several QTLs. Thus, different BC4F3:5 lines were obtained from the BC3DH lines by backcrossing and self-pollination (Fig. 3b), after which homozygosity was checked using the SSR markers carried by the introgressed Og_MG12 segments present in the BC3DHs. Panicle traits were phenotyped for five BC4F3:5 lines (namely L_A to L_E) containing the different individual O. glaberrima introgressed segments affecting TBN, together with Os_Caiapó and Og_MG12 (Fig. 3c; Additional file 1: Table S1).

Divergent panicle traits compared to the Os_Caiapó parent were observed in all BC4 lines except for L_C (Fig. 3b). The latter contains an introgression at the beginning of q_7 (from RM11 to RM10), suggesting that this region does not influence panicle architecture. All the other BC4 lines showed longer primary branches compared to the Os_Caiapó parent, suggesting that introgressions in chromosomes 1, 3 and 7 could independently influence primary branch length (Fig. 3c). The L_A line, containing Og_MG12 regions corresponding to q_1, q_2 and q_3-2, additionally showed a decreased PBN value associated with a higher SBN/PBN ratio. This line did not produce any tertiary panicle branch. This suggests that the association of the q_1, q_2 and q_3-2 regions could cause a reduction in PBN associated with an increased SBN/PBN ratio.

Line L_B, containing only one Og_MG12 introgression (from RM60 to RM7) in place of q_3-1, showed increased values for four panicle traits (PBN, PBL, SBN/PBN and SpN) but not for TBN. However, these results were not observed during the second phenotyping of this line, except for an increase in PBL (Additional file 2: Figure S4). Thus, it can be deduced that the q_3-1 region positively influences primary branch length in a manner that is independent of the other detected QTLs and that additive panicle trait effects may be observed depending on the environment.

Finally, the BC4 lines L_D and L_E were the only ones observed to develop tertiary branches (Fig. 3c). In both lines, this phenotype is associated with an increased SBN/PBN and a longer PBL. The higher SBN/PBN ratio is associated with a decreased PBN in L_D, meaning that in this line more secondary branch meristems are established during panicle development. Similar results for SBN/PBN and TBN were observed in a second round of phenotyping, with a comparable decrease in PBN in the L_E line (Additional file 2: Figure S4).

The two aforementioned lines contain contiguous introgression fragments in chromosome 7, from RM10 to RM18 for L_D and from RM134 to RM118 for L_E. Since the exact recombination positions of the introgressed segments for each BC4 is not known, the regions between SSRs with different genotypes are included in the QTL intervals. In Fig. 4a, it can be observed that lines L_D and L_E, which share the same phenotypes except for PBN, may harbor introgressions that overlap over only a small region between RM18 and RM134. Based on these observations, two different hypotheses can be proposed with regard to QTL position(s) in the q_7 region (Fig. 4a). The first one postulates the existence of a common QTL located between RM18 and RM134 that controls PBL, SBN/PBN and TBN. The second hypothesis is that two different QTLs, q_7-1 between RM10 and RM134, and q_7-2 between RM18 and RM420, exert a similar effect on PBL, SBN/PBN and TBN but act differently on PBN.

Fig. 4
figure 4

Description of genetic variations observed in the q_7 region between the Os_Caiapó and Og_MG12 genomes (a) Influence of the q_7 region on panicle variation in the BC4 lines. Light green and green shading indicate the “extended” and strict introgression positions respectively for each BC4 line. b Representation of variation observed in the annotated genes between the Os_Caiapó and Og_MG12 genomes in the region between the RM10 and RM420 SSR markers. Yellow bars show loci only present in Os_Caiapó genome, green bars indicate loci only present in Og_MG12. Light green bars represent genes that are duplicated in Og_MG12 and pink bars represent genes that are annotated in Og_MG12 but present in a different genomic region in Os_Caiapó. c Schematic representation of sequence divergence in candidate genes between the two genomes. Blue lines correspond to promoter regions, red and grey boxes represent exons and introns respectively. SNP and InDel variations leading to TFBS or amino acid changes are represented by dots and triangles respectively. The number above a triangle indicates InDel size. Asterisks indicate amino acid identity modification in the protein sequence. TFBSs present only in Os_Caiapó or Og_MG12 are shown colored in yellow and green respectively. TFBS: Transcription Factor Binding Site

Taken together, the above results suggest that panicle architecture is a complex trait controlled by various different genomic regions which positively or negatively influence branching. An association of q_1, q_2 and q_3-2 produces opposing effects on PBN and SBN while only the q_3-1 region, comprised between RM175 and RM7, may impact positively upon PB length. Finally, the region in q_7, which negatively influences the number of primary branches and may be associated with the formation of tertiary branches and an increase in SBN per primary branch, may be defined as being either between the RM18 and RM134 marker positions or associated with two QTLs positioned between RM10 and RM134 for q_7-1 and from RM18 to RM420 for q_7-2.

Co-location of QTLs with known genes and with QTLs detected in other populations

Many studies have reported QTLs associated with panicle architecture and/or yield traits or both, using bi-parental (QTL mapping) or diversity panels GWAS [25, 35, 41,42,43,44,45,46,47,48,49]. We found 76 common sites between the QTLs detected in this study and those reported earlier (Additional file 1: Table S5). Among them, five sites had been previously described in the evaluation of a CSSL population between O. sativa and O. glaberrima with respect to traits affecting panicle structure and grain yield [41]. In addition, co-located sites corresponding to similar traits (PBN, SBN and SpN) were found on chromosomes 11 and 12 through studies of various different rice populations [35, 43,44,45, 49]. Although the common sites identified all relate in some way to panicle architecture or yield, care should be taken when attempting to extrapolate between the different studies due to variations in the methodologies used to record traits.

To evaluate the synteny of the QTL regions between the genomes of the two CSSL parents, sequences from the assemblies of the Os_Caiapó and Og_MG12 genomes obtained by ONT sequencing (;; Additional file 1: Table S6) were used to carry out local alignment and structural variation analyses for each QTL region. The assembled genomes are of high continuity and completeness, with BUSCO score of 98.4% allowing precise detection of structural variations. For the majority of the QTL regions, no major structural variation was observed between the two genomes (Additional file 2: Figure S5). However, some notable differences were detected at the qPBN2, qTBN5, qTBN7, qSpN3 and qSBN12/PBN12 sites. Within QTLs qPBN2 and qTBN7, segments of respective lengths 1,296,899 and 230,762 bp in the Os_Caiapó genome were inverted in the Og_MG12 genome (Additional file 1: Table S7). The genomic regions for QTLs qTBN5, qSpN3 and qSBN/PBN12 displayed large InDels between Os_Caiapó and Og_MG12 (Additional file 1: Table S7). With the exception of qSBN11, we observed a high sequence similarity in the QTL regions between the two genomes, suggesting similar content in terms of coding sequences. Synteny analysis was extended by the alignment of the Os_Caiapó and Og_MG12 QTL regions with those of the O. sativa cv. Nipponbare reference genome (Additional file 2: Figure S5). Based on the high conservation of the corresponding regions between Os_Caiapó and Os_Nipponbare, we then used as a reference, for subsequent analyses, the O. sativa cv. Nipponbare (IRGSPv1.0) functional gene annotation databases (i.e., MSU7 and RAP_db) for candidate gene prioritization in relation to panicle architecture and flowering.

For all QTLs identified in the present study, we found several known genes with relevant functions that related to flowering and/or panicle development (Additional file 1: Table S8; Additional file 2: Figure S6). Genes such as LAX PANICLE1 (LAX1), OsMADS1/LEAFY HULL STERILE1 (LHS1), OsMADS14, OsMADS34/PANICLE PHYTOMER2 (PAP2), OsINDETERMINATE SPIKELET 1 (OsIDS1), OsMADS18, DENSE AND ERECT PANICLE2 (DEP2) and FRIZZY PANICLE (FZP) are known to be involved in the control of panicle development and were suggested as candidates that might contribute to the panicle branching diversity observed between the two parents Os_Caiapó and Og_MG12 [3,4,5].

q_7 genetic variations between Os_Caiapó and Og_MG12

We paid particular attention to the q_7 region, as it was found to be associated with several panicle morphological traits and to span a genomic region that included several panicle-associated genes. To test the hypothesis that specific genetic modifications present in the q_7 QTL region could be associated with variation in panicle architecture, we further compared the constituent genes in the region between the RM10 and RM420 markers in the Os_Caiapó and Og_MG12 genomes on chromosome 7 (22.23 – 29.59 Mbp in Os_Caiapó vs. 20.12 – 26.69 Mbp in Og_MG12). For this purpose, gene annotation comparisons and BLAST analyses were performed to explore in detail gene synteny and presence/absence of genes between the Os_Caiapó and Og_MG12 genomes within this region (Fig. 4b; Additional file 1: Table S11). Between the RM10 and RM420 marker positions, we observed a variation in the number of annotated genes between the two genomes (Additional file 1: Table S9). This variation is due to several factors: the differential presence of corresponding orthologs between the two genomes; differential gene duplication within the region; and different locations of a given gene within the two genomes (i.e., genomic rearrangement) (Fig. 4b; Additional file 1: Table S9). Among the genes that differ between the two genomes in this region, many are similar to transposable elements (TEs) or are hypothetical protein-encoding genes. None of them has a known function related to panicle development or control of flowering, or to any other developmental processes.

Special attention was paid to common genes present in the q_7 region that were known to be related to inflorescence development and/or meristem activity and maintenance. Based on a bibliographic survey, no candidate gene of special interest was identified within the region between the RM18 and RM134 markers. In contrast, several interesting candidate genes were identified in the q_7-1 and q_7-2 regions: the OsHOX14, OsMADS18 and DEP2 genes in q_7-1 and the FZP and WOX11 genes in q_7-2 (Fig. 4c). SNP and InDel sites associated with the aforementioned genes that were polymorphic between the Os_Caiapó and Og_MG12 genomes were analyzed in order to detect amino acid modifications, open reading frame alterations or transcription factor binding site (TFBS) variations within promoter regions.

OsHOX14 (LOC_Os07g39320/Os07g0581700) encodes a protein of the Homeodomain-leucine zipper (HD-Zip) TF family [50, 51] and two of the identified SNPs cause amino acid variations in the coding sequence, including one in the homeobox domain (Fig. 4c; Additional file 1: Table S10). Comparison of the OsHOX14 and OgHOX14 promoter regions revealed InDel and SNP variations leading to the loss of nine TFBSs and a gain of three new TFBSs in the CSSLs bearing Og_MG12 q_7-1 introgression (Fig. 4c; Additional file 1: Table S11).

OsMADS18 (LOC_Os07g41370/Os07g0605200) encodes a protein of the AP1/FUL-like MADS-box TF subfamily [52, 53]. SNPs were detected in the coding sequence; one of the SNPs leads to a non-synonymous change in the OgMADS18 protein outside the known binding or functional domains (Fig. 4c; Additional file 1: Table S10). Scanning of the promoter regions of OsMADS18 and OgMADS18 in the Os_Caiapó and Og_MG12 genomes respectively revealed several SNPs and InDels that lead to variations between the two genomes (presence/absence) of TFBSs recognized by AP2, NAC or HB TFs (Fig. 4c, Additional file 1: Table S11).

DEP2 (LOC_Os07g42410/Os07g061600) encodes a plant-specific protein of unknown function [54]. Various SNPs and one InDel are observed between the coding sequences of OsDEP2 and OgDEP2 which result respectively in changed amino acid identities between the two protein orthologs (Fig. 4c; Table S9) and an insertion of two amino acids in the OgDEP2 protein. Sequence variations were also observed between the promoters of OsDEP2 and OgDEP2; these variations lead to the loss of bZIP, B3, AT-Hook and AP2 TFBSs in the CSSLs carrying q_7. On the other hand, in comparison to O. sativa, the CSSLs carrying q_7 include NAC and bZIP TFBSs specific to the promoter of OgDEP2.

Among the genes contained in the q_7-2 region, FZP (LOC_Os07g47330/Os07g0669500) encodes an AP2/ERF TF [20, 55]. Several SNP and InDel variations observed in the coding region of FZP lead to coding sequence amino acid changes between the two genomes (Fig. 4c; Additional file 1: Table S9). An InDel of 9 bp in OgFZP results in the insertion of three histidine amino acids outside the single AP2 domain in the OgFZP protein. We observed several SNPs and InDels between the promoter of the two orthologs, which result in a differential presence of TFBSs (Fig. 4c, Additional file 1: Table S10). No variation was observed in the copy number variant (CNV) motif of 18 bp described by Bai et al. (2017) [18]. Huang et al. (2018) [19] revealed a deletion of 4 bp in the OsFZP 5' regulatory region in comparison to the sequence of O. rufipogon. This deletion is observed in Os_Caiapó and not in the genome of Og_MG12 (Fig. 4c). Recently, it has been shown that CU-rich elements (CUREs) present in the 3' UTR of the OsFZP mRNA are crucial for efficient OsFZP translational repression [56]. Three CUREs are detected in the 3' UTR of OsFZP in the Os_Caiapó genome. In contrast, a deletion of the third CURE sequence is observed in the Og_MG12 genome (Fig. 4c).

In the WUSCHEL-related homeobox WOX11 gene (LOC_Os07g48560/Os07g0684900), an InDel in the coding region of WOX11 leads to an Asn amino acid insertion in the OgWOX11 protein in Og_MG12 without affecting the homeobox domain of the protein (Fig. 4c, Table S10). In the promoter region, variations between Os_Caiapó and Og_MG12 lead to the presence/absence of several TFBSs in the promoter of OgWOX11 (Fig. 4c; Table S11). We also observed a large insertion of 3,760 bp, containing about 31 AP2 TFBs, in Og_MG12 compared to the Os_Caiapó genome.

Overall, our analysis revealed variations in synteny that demonstrate the absence of certain Os_Caiapó genes and the addition of other loci as a consequence of Og_MG12 genomic introgressions within the CSSLs: some of these changes can be hypothesized to play a role in determining panicle trait diversity. Moreover, our analysis of candidate genes revealed TFBS variations and protein coding sequence polymorphisms that may lead to variations in transcript levels and/or protein activity in CSSLs harboring the q_7 introgression.


Added value of the O. sativa x O. glaberrima CSSL population

Interspecific O. sativa x O. glaberrima CSSL genetic resources are of great interest: O. glaberrima provides a gene pool with high potential for rice improvement in terms of resistance to biotic and abiotic stresses and ecological adaptability [40, 41]. CSSL libraries are also good pre-breeding materials for the simultaneous identification, transfer and pyramiding of key genes in crop improvement programs. Genetic effects resulting in changes to panicle architecture could potentially be obtained via genome editing as an alternative to pyramiding QTLS producing the same effect. They are also useful for the study of traits lost or retained during the process of evolution, domestication and breeding [38, 57]. Genetic strategies based on CSSL populations have the advantage of using a relatively small number of lines for experiments, allowing replicating evaluations, thereby enhancing statistical strength for complex and time-consuming phenotypic assessments.

In this study, backcross introgression lines (BC3DH) of O. sativa_Caiapó x O. glaberrima_MG12 were phenotyped for panicle morphological traits, which are key components of yield potential. Our analyses led to the detection of 15 QTLs, localized on chromosomes 1, 2, 3, 5 and 7 and comprising several genes known to be involved in panicle architecture determination (Table 2). Colocalization was observed with previous QTLs and GWAS sites associated with panicle branching traits, supporting the implication of the QTLs detected in this study in the regulation of panicle morphology. Moreover, since the genomic sites mentioned above were detected using different sets of populations, we propose that the genomic regions in question play a major role in the determination of panicle architecture diversity.

For most of the traits studied, we detected several QTLs within the same line, as observed in the lines L_10 and L_46 that allowed the localization of QTLs governing TBN-, PBN- and PBL-related traits. This is not surprising in the case of a complex trait such as panicle architecture, which is controlled by a large number of genes and QTLs with small effects that can be influenced by environmental and epistatic interactions [58]. The fact that we observed a colocalization of QTLs for different panicle traits suggests a likely genetic interdependence between these traits; this hypothesis is supported by correlations observed between phenotypic values for morphological characters such as SpN, SBN, PBN, PBL and RL that share common QTLs.

Impact of O. glaberrima introgressions on panicle traits in the O. sativa background

Compared with the Os_Caiapó recurrent parent, the Og_MG12 donor parent displays a global reduction in the number of panicle constituents (i.e., numbers of primary branches, secondary branches and spikelets) and produces longer primary branches. Tertiary branches were not observed in the Og_MG12 donor parent. On the basis of these observations, the QTLs detected in the CSSLs, as expected, were associated with a global decrease in panicle constituent numbers (PBN, SBN, SpN) and an increase of PB length in comparison to the Os_Caiapó parent. In most of the lines, introgressions of Og_MG12 alleles of panicle-regulating genes produced a negative effect on the regulatory network controlling panicle morphological traits, indicating that some of the genomic variations observed between Og_MG12 and Os_Caiapó in these regions are functionally important.

Surprisingly, we also detected QTLs associated with an increase in panicle trait values in comparison to the Os_Caiapó recurrent parent. For example, the Og_MG12 introgression localized on chromosome 3 in place of q_3-1 induces a globally higher complexity of panicle branching with an increase in PBN, SBN/PBN and SpN, associated with an increased PBL. This region includes the gene OsMADS1/LHS1 belonging to the SEPALLATA MADS-Box TF family, which is known to play an important role in determining floral meristem identity and in floral organ development [59,60,61]. In O. sativa, OsMADS1/LHS1 directly regulates several other transcription factor genes (MADS box and Homeodomain family members) and hormone signaling pathways implicated in floral meristem specification, maintenance and determinacy [62]. These unexpected effects indicate that introgression of certain Og_MG12 regions into the Os_Caiapó genetic background can lead to an enhanced panicle phenotype and suggest a perturbation of the native pathways regulating inflorescence development in Os_Caiapó. This perturbation could be explained by several mechanisms. Firstly, it is possible to postulate the presence of a negative regulator of branching in Os_Caiapó that is absent in the Og_MG12 introgression, leading to an upregulation of branching in the CSSL. A second hypothesis could be proposed whereby one or more genes specific to O. glaberrima are present in the Og_MG12 introgression compared to the corresponding Os_Caiapó region. One or more of these genes could be positive regulators of panicle architecture, with this regulatory potential being either latent or only very weakly expressed in the Og_MG12 background. A third possible explanation could be proposed, involving cis-variations associated with orthologous genes conserved between the two genomes. In this case, the addition of new Og_MG12 genes, or their variant allelic forms into the Os_Caiapó genomic background could be postulated to exert an additive effect on panicle branching in the CSSLs, by affecting regulatory interactions within the Os_Caiapó network governing panicle branching.

In a similar way, our observations on the effects of QTL q_7, which favored increased SB and TB numbers, might have more than one possible explanation. Comparative analysis of the BC4 lines led us to suggest two hypotheses (above) regarding the position of QTL(s) in the q_7 region (Fig. 4a). However, based on quantitative and qualitative differences between the BC4 line phenotypes, we favor the hypothesis whereby two distinct regions influence panicle trait variation differentially (i.e., the q_7-1 and q_7-2 regions). In this scenario, both would be associated with increased PBL, SBN and TBN traits values whereas q_7-1 would have an additional negative effect on PBN. Moreover, the overlapping region of the introgressed segments in BC4 lines L_D and L_E, which spans a region of ~ 982 Kbp, is recombined in both lines. This implies that lines L_D and L_E can bear the Os allele at any locus in this region, so the probability that both lines bear the Og allele at a same putative QTL position is low. In addition, this region only contains a few genes whose predicted function is unknown.

In parallel, we observed a good synteny between the two species within the q_7-1 and q_7-2 regions, ruling out the possibility that large-scale genomic rearrangements might explain the phenotypic variations observed. Indeed, the strong global synteny between these regions suggests rather that CSSL phenotypes might be accounted for by modifications within the genomic regions of key regulatory genes that give rise to variations in their expression or in their protein functions and/or interactions through gene regulatory networks (GRNs). Mutations affecting transcription factor proteins and/or DNA sequences or both, such as enhancers and promoters, fall into this category [63]. Moreover, traits associated with dynamic processes are more readily modified through their cis- and/or trans-regulation rather than through coding sequence mutations [64]. The q_7-1 and q_7-2 regions comprise several genes known for their role in panicle development and numerous SNPs and InDels have been detected in the promoter regions and coding sequences of these candidate genes.

Within the q_7-1 region, three genes have been reported to have a direct function in panicle architecture establishment. The homeodomain-leucine zipper transcription factor gene OsHOX14 has been shown to be involved in the regulation of panicle development and its loss of function leads to a reduction in PBN compared to wild type [50, 51]. The second candidate gene OsMADS18, in addition to its role in flowering time promotion [53], is known to specify inflorescence meristem identity through interaction with PANICLE PHYTOMER2 (PAP2/OsMADS34) and the two other members of the AP1/FUL subfamily, OsMADS14 and OsMADS15 [14]. DEP2 encodes a plant-specific protein without a known functional domain, and some of its mutations affect elongation of the rachis and both primary and secondary branches, caused by a defect in cell proliferation [54, 65, 66]. For each of these three genes, the Og_MG12 allelic version shows variations in comparison to Os_Caiapó in both protein coding sequences and in putative TFBSs within the promoter region. The presence of these binding sites may reveal a divergence in the regulation of the expression of these genes between the two parents, which could lead to the morphological variations observed in the CSSLs (i.e., decrease in PBN).

Concerning the q_7-2 region associated with the formation of higher order panicle branches (i.e., secondary and tertiary branches), the candidate genes WOX11 and FZP were detected in both genomes. The WUSCHEL-related homeobox gene OsWOX11 is necessary and sufficient to promote crown root emergence and growth in rice by either responding to or by regulating auxin and cytokinin signaling [67,68,69,70,71,72]. Recently, Cheng et al. (2018) [73] showed that OsWOX11 and JMJ705 cooperatively control shoot growth and commonly regulate the expression of a set of genes involved in meristem identity, indicating that the role of this gene is not confined to roots. A comparison of the promoter regions of WOX11 orthologs revealed numerous SNP and InDel differences between the two genomes and an insertion of 3,760 bp in Og_MG12, which includes numerous AP2 TFBS potentially recognized by the MULTI-FLORET SPIKELET 1 (MFS1) and OsFZP genes, known to play an important role in the regulation of spikelet meristem determinacy [12, 20, 55, 74]. In an O. sativa genomic environment, the promoter of OgWOX11 containing new AP2 TFBSs could potentially be programmed differently in its expression during the reproductive phase of the plant due to altered promoter activity, with possibly effects on cell proliferation in the axillary meristem leading to higher order branching in the CSSL lines.

The OsFZP gene plays an important role in the transition from branch to spikelet primordium during panicle development [12, 20, 55]. Reduced expression of OsFZP at the reproductive stage increases the extent of higher order branching of the panicle, resulting in increased grain number [24]. Recent studies showed that fine tuning of OsFZP expression at the transcriptional and post transcriptional levels could affect panicle architecture [18, 19, 21, 56, 75]. Some SNPs and InDel variations between the Os_Caiapó and Og_MG12 genomes were detected within the regulatory regions of the respective FZP orthologs, with notable polymorphic sites occurring close to an auxin response element (AuxRE) 2.7 Kpb upstream of OsFZP and within the CU rich elements (CURES) localized in the 3'UTR of the same gene [19, 56]. These two variations observed in regulatory regions of OgFZP could result in an increase in transcript and/or protein amounts, leading to a precocious transition from branch to spikelet primordium. Such a scenario would be corroborated by the lesser-branched panicle phenotype of Og_MG12 compared to Os_Caiapó but not by the phenotype of the CSSLs containing OgFZP. The analysis of the combinatorial effect of variations in these regulatory elements would be useful to explain the observed phenotypes.


This study was carried out with the broad aim of identifying discrete genetic elements that govern rice panicle architectural diversity using a population of interspecific introgression lines. We detected several QTLs associated with rice yield and panicle branching diversity observed between the two cultivated rice species O. sativa and O. glaberrima, with decreased values – as expected – in the CSSLs but also unexpectedly with added values for some QTLs compared to the O. sativa recurrent parent. This was the case for QTL q_7 on chromosome 7, which causes an increase in numbers of higher order panicle branches attributable to two distinct genomic regions: q_7-1 and q_7-2. A detailed comparative genomic analysis revealed variations in gene content, promoter region sequence and protein coding sequences for a number of different key genes that act in the control of panicle architecture.

When an Og_MG12 allelic variant is placed in an Os_Caiapó genomic background context within a CSSL, it might in some instances cause increased panicle branching by acting in an additive fashion, through enhancement of the native Os_Caiapó gene regulatory network (GRN) that governs panicle development. Importantly, the biological functions and regulation of a number of developmentally important genes present in the q_7-1 and q_7-2 regions have already been described in O. sativa; however, functional analyses of their O. glaberrima orthologs have yet to be undertaken in order to confirm the conservation of biological roles and/or to assess their regulation. Further studies assessing how the genes of interest are regulated will be beneficial to the improvement of rice breeding strategies for the exploitation of favorable alleles present in this CSSL population. The genetic material created here will provide knowledge and resources to facilitate the retention of favorable alleles in breeding programs and to develop new improved rice cultivars using O. glaberrima as a genetic resource for enhancing morphological traits in O. sativa.

Materials and methods

Plant materials

The plant materials consisted in a set of 60 interspecific introgression lines, or Chromosome Segment Substitution Lines (CSSL). The CSSLs were derived at CIAT, Cali, Colombia from a cross between O. sativa subgroup tropical japonica (cv. Caiapó) as the recurrent parent and O. glaberrima (cv. MG12; acc. IRGC103544) as the donor parent [40]. The CSSLs used were BC3DH, that is, obtained after three rounds of backcrossing to the recurrent parent followed by double haploidization of BC3F1 male gametes (Additional file 1: Table S1). They were genotyped at CIAT using 200 simple-sequence repeat (SSR) markers [40]. The program CSSL Finder v1.0.0 ( [40] was then used to select the 60 lines from an initial panel of 312 BC3DH lines, so that the entire donor genome was represented by overlapping chromosome segments. The effects of O. glaberrima introgressions on panicle trait variation were evaluated in five BC4F3:5 lines, derived so that they contained the target region and as few introgressed O. glaberrima genomic segments in the rest of the genome as possible (Additional file 1: Table S1).

Phenotypic evaluation

The 60 BC3DH lines, along with the parents (Os_Caiapó and Og_MG12), were grown together in November 2012-January 2013 (Year 1) in the International Center for Tropical Agriculture (CIAT, now Alliance Bioversity-CIAT) headquarters experimental fields (Palmira, Colombia) (76° 21'W, 3° 30'N and 967masl) under irrigated conditions. The experimental design was a randomized complete block design with two replications and 62 plots. Five plants per line per plot were grown. Panicle traits were evaluated for subsequent QTL analysis and validation. For each line, the three main panicles from three randomly chosen plants (9 panicles total) per line per replicate were collected. Each panicle was spread out on a white background and held in place with metal pins for photography shooting. A total of 1,443 panicles from the 60 BC3DH lines and the parents were dissected and scored manually. In January-April 2014 (Year 2), a total of nine BC3DH lines were re-evaluated using P-TRAP software [76], which performs automatic scoring of panicle traits. Five BC4F3/F5 lines with parents, plus L_10 and L_46 BC3DH, were grown in the greenhouse in Montpellier, France (March 2021) at 28 °C / 80% relative humidity under short days conditions (11h light-13h dark), and were scored for panicle traits using P-TRAP. Panicle phenotypes were re-evaluated in the same conditions for three BC4 lines (L_B, L_D and L_E in June 2021).

For each analysis (2013, 2014 and 2021), 6 morphological panicle traits were scored: rachis length (RL); number of primary branches (PBN); number of secondary branches (SBN); number of tertiary branches (TBN); spikelet number (SpN); and average primary branch length (PBL) per panicle (Additional file 1: Table S2).

Statistical analysis and QTL detection

To check for genotypic and genotype \(\times\) environment effects, analysis of variance (ANOVA) was performed on each trait in the R software package (version 4.0.4, R Core Team 2022), taking into consideration replicates and lines as fixed effects:

$${y}_{ij}=\mu +{g}_{i}+{r}_{j}+{\varepsilon }_{ij}$$

where \({y}_{ij}\) is the phenotype score of the line \(i\) in repetition \(j\), \(\mu\) is the average of phenotypic values across the population, \({g}_{i}\) is the random effect of genotype \(i\), \({r}_{j}\) is the random effect of repetition \(j\) and \({e}_{ij}\) is a residual error associated with line \(i\) and repetition \(j\). The ANOVA results showed that lines differed significantly for all traits (p-value \(<0.001\)) (Additional file 1: Table S3), indicating strong genotypic effects. The factor "repetition" was also significant for the RL, PBL and traits, although it was much less significant than the genotype effect (Additional file 1: Table S3). Thus, we decided to carry a preliminary QTL analysis – with the methodology described hereby – on each repetition separately. As a result, the F-test profiles were much similar between the two repetitions (Additional file 2: Figure S1). Therefore, the subsequent analyses presented in this work were done taking the average phenotypic values of the two repetitions.

Trait heritability was computed using the line effect based on the variance among phenotypic measurements between the two replicates of the population phenotyping assay. The corrplot (function cor), ade4 (function dudi.pca using center = TRUE and Scale = TRUE as arguments) and devtool R packages were used to analyze phenotypic correlation between traits.

The CSSL Finder program was used to display graphical genotypes of the 200 SSRs in the 60 BC3DH CSSLs together with phenotypic values for each trait. Putative QTLs were detected using the graphical genotyping tool proposed by CSSL Finder, which allows combination of logical genotype–phenotype association with single-marker ANOVA1 F-test. The F-test was considered significant when its value was higher than 10.0, which corresponds to a \(p-value<\sim 0.002\) considering the degrees of freedom of our experimental design.

Because the F-test can lead to detection of false QTLs when the numbers of lines with and without an introgression in the considered region, putative QTLs were submitted to a comparison between the CSSLs and the recurrent parent Os_Caiapó was performed by a Dunnett's multiple comparison test (\(p<0.05\)) (glht function in R). The relative effect RE of an O. glaberrima introgression in a given CSSL was calculated using the least square means (LSMEANS) output of the GLM procedure as follows:

$$RE=100\times \left[LsMean\left(Lines\right)-LSMean\left(Os\_Caiap{\acute{o} }\right)\right]/LSMean\left(Os\_Caiap{\acute{o} }\right)$$

A putative QTL detected by the F-test was considered as validated when the trait values of the lines bearing the introgression at the QTL location were significantly different—according to the Dunnett's test—from that of the parent Os_Caiapó \((p<0.001)\) and when at least two lines bearing overlapping segments shared similar phenotypes. The phenotypic variation of these lines was further evaluated in the second phenotyping experiment in Year 2.

Genome sequencing, assembly, quality assessment and annotation

Samples of young leaves from O. sativa cv. Caiapó and O. glaberrima cv. MG12 were used to extract genomic DNA and for preparation of SKL-LSK109 libraries as described in [77]. Genome sequencing was performed on a Nanopore MinION Flow Cell R9.4.1 (Oxford technologies, Oxford Science Park, UK). The long-read sequences of Os_Caiapó and Og_MG12 generated by the ONT sequencing were assembled using the following steps. After basecalling and Q > 8 filtering with Guppy V6.1.2, with the SUP model, the preliminary genome was assembled with Flye V2.9 and default assembly parameters for Sup ONT [78]. Medaka V1.6.1 was then used to create consensus sequences and to improve the accuracy of the assembly (polishing). The contigs were ordered with RagTag V2.1.0 using O.sativa cv. Nipponbare/IRGSP1.0 as a reference and with default parameters [79]. An evaluation of genome assembly completeness was carried out using BUSCO (Benchmark Universal Single Copy Orthologs) v5.2.2 with default parameters and the Poale database (Additional file 1: Table S7) [80].

The Liftoff tool was used for gene prediction and annotation using the genome reference O. sativa cv. Nipponbare (IRGSPv1.0, MSU7.0) and associated gff files for both Os_Caiapó and Og_MG12 and by using the O. glaberrima genome reference OglaRS2 along with its associated gff files ( in the case of Og_MG12 (Additional file 1: Table S7) [81].

Genomic alignment

The sequences of the QTL regions and the genes of interest were extracted from the three genomes O. sativa cv. Caiapó, O. glaberrima cv. MG12 and O. sativa cv. Nipponbare MSU7 using Seqkit version 2.4.0 [82]. Sequences of QTL regions were aligned and plotted using the minimap2 aligner implemented in the D-GENIES online tool ( using the option Few repeats.

QTL colocalization, identification and bio-analysis of candidate genes

For QTL and gene colocalization, the qTARO database (, the funRiceGenes database ( [83] and various datasets from recently published works were used to identify QTLs and genes that overlapped with the QTL regions detected in this study. To identify genes potentially associated with the QTL regions, we used the O. sativa cv. Nipponbare MSU7.0 ( and the RAP_db ( annotation databases. From the annotated gene list, the candidate genes were identified based on their predicted function (biological processes), their referencing in the funRiceGenes database and/or their expression pattern with respect to the trait of interest [34, 84]. Regions spanning the candidate genes and their associated 5 Kbp upstream region (considered as their promoter regions) were obtained for O. sativa (IRGSPv1.0, MSU7.0) and for O. glaberrima (OglaRS2), then nucleotide and protein alignments were performed using CLUSTALW in order to identify SNPs, InDels and amino acid changes between the two species [85]. The Plant Promoter Analysis Navigator web facility ( was used to detect transcription factor binding sites (TFBSs) and regulatory elements (CpG islands and tandem repeats), with a cut-off at 0.8, in the promoter region of candidate genes [86]. Variations between the two genomes, in terms of promoter and gene structures along with SNP and InDel positions and information, were drawn using the R package KaryoploteR [87].

Availability of data and materials

The CSSLs materials are available from the corresponding authors on reasonable request. Sequencing data generated for this project are available at and for Os_Caiapó and Og_MG12 respectively.



Rachis Length


Primary Branch Number


Primary Branch Length


Secondary Branch Number


Spikelet Number


Ratio of SBN per PBN

Os_Caiapó :

Oryza sativa Cv. Caiapó;

Og_MG12 :

Oryza glaberrima Cv. MG12


Chromosome Segment Substitution Lines




Simple Sequence Repeat


Transcription Factor Binding Site


  1. Itoh JI, Nonomura KI, Ikeda K, Yamaki S, Inukai Y, Yamagishi H, et al. Rice plant development: from zygote to spikelet. Plant Cell Physiol. 2005;46:23–47.

    CAS  PubMed  Google Scholar 

  2. Xing Y, Zhang Q. Genetic and molecular bases of rice yield. Annu Rev Plant Biol. 2010;61:421–42.

    CAS  PubMed  Google Scholar 

  3. Chongloi GL, Prakash S, Vijayraghavan U. Regulation of meristem maintenance and organ identity during rice reproductive development. J Exp Bot. 2019;70:1719–36.

    CAS  PubMed  Google Scholar 

  4. Yin C, Zhu Y, Li X, Lin Y. Molecular and Genetic aspects of grain number determination in rice (Oryza sativa L.). Int J Mol Sci. 2021;22:728.

    CAS  PubMed  PubMed Central  Google Scholar 

  5. Wang C, Yang X, Li G. Molecular insights into inflorescence meristem specification for yield potential in cereal crops. Int J Mol Sci. 2021;22:3508.

    CAS  PubMed  PubMed Central  Google Scholar 

  6. Lu Y, Chuan M, Wang H, Chen R, Tao T, Zhou Y, et al. Genetic and molecular factors in determining grain number per panicle of rice. Front Plant Sci. 2022;13:964246.

    PubMed  PubMed Central  Google Scholar 

  7. Chun Y, Kumar A, Li X. Genetic and molecular pathways controlling rice inflorescence architecture. Front Plant Sci. 2022;13:1010138.

    PubMed  PubMed Central  Google Scholar 

  8. Tanaka W, Yamauchi T, Tsuda K. Genetic basis controlling rice plant architecture and its modification for breeding. Breeding Sci. 2023;73:3–45.

    Google Scholar 

  9. Ashikari M, Sakakibara H, Lin S, Yamamoto T, Takashi T, Nishimura A, et al. Cytokinin oxidase regulates rice grain production. Science. 2005;309:741–5.

    CAS  PubMed  Google Scholar 

  10. Jiao Y, Wang Y, Xue D, Wang J, Yan M, Liu G, et al. Regulation of OsSPL14 by OsmiR156 defines ideal plant architecture in rice. Nat Genet. 2010;42:541–4.

    CAS  PubMed  Google Scholar 

  11. Miura K, Ikeda M, Matsubara A, Song X-J, Ito M, Asano K, et al. OsSPL14 promotes panicle branching and higher grain productivity in rice. Nat Genet. 2010;42:545–9.

    CAS  PubMed  Google Scholar 

  12. Komatsu M, Maekawa M, Shimamoto K, Kyozuka J. The LAX1 and FRIZZY PANICLE 2 genes determine the inflorescence architecture of rice by controlling rachis-branch and spikelet development. Dev Biol. 2001;231:10–10.

    Google Scholar 

  13. Kobayashi K, Maekawa M, Miyao A, Hirochika H, Kyozuka J. PANICLE PHYTOMER2 (PAP2), encoding a SEPALLATA subfamily MADS-box protein, positively controls spikelet meristem identity in rice. Plant Cell Physiol. 2009;51:47–57.

    PubMed  PubMed Central  Google Scholar 

  14. Kobayashi K, Yasuno N, Sato Y, Yoda M, Yamazaki R, Kimizu M, et al. Inflorescence meristem identity in rice is specified by overlapping functions of three AP1/FUL-Like MADS Box Genes and PAP2, a SEPALLATA MADS Box Gene. THE PLANT CELL ONLINE. 2012;24:1848–59.

    CAS  Google Scholar 

  15. Huang XX, Qian QQ, Liu ZZ, Sun HH, He SS, Luo DD, et al. Natural variation at the DEP1 locus enhances grain yield in rice. Audio Transact IRE Professional Group On. 2009;41:494–7.

    CAS  Google Scholar 

  16. Li X, Qian Q, Fu Z, Wang Y, Xiong G, Zeng D, et al. Control of tillering in rice. Nature. 2003;422:618–21.

    CAS  PubMed  Google Scholar 

  17. Yoshida A, Sasao M, Yasuno N, Takagi K, Daimon Y, Chen R, et al. TAWAWA1, a regulator of rice inflorescence architecture, functions through the suppression of meristem phase transition. Proc National Acad Sci. 2013;110:767–72.

    CAS  Google Scholar 

  18. Bai X, Huang Y, Hu Y, Liu H, Zhang B, Smaczniak C, et al. Duplication of an upstream silencer of FZP increases grain yield in rice. Nat Plants. 2017.

    Article  PubMed  Google Scholar 

  19. Huang Y, Zhao S, Fu Y, Sun H, Ma X, Tan L, et al. Variation in the regulatory region of FZP causes increases in secondary inflorescence branching and grain yield in rice domestication. Plant J. 2018;96:716–33.

    CAS  PubMed  Google Scholar 

  20. Komatsu M, Chujo A, Nagato Y, Shimamoto K, Kyozuka J. FRIZZY PANICLE is required to prevent the formation of axillary meristems and to establish floral meristem identity in rice spikelets. Development. 2003;130:3841–50.

    CAS  PubMed  Google Scholar 

  21. Fujishiro Y, Agata A, Ota S, Ishihara R, Takeda Y, Kunishima T, et al. Comprehensive panicle phenotyping reveals that qSrn7/FZP influences higher-order branching. Sci Rep-uk. 2018;8:12511.

    Google Scholar 

  22. Ikeda K, Nagasawa N, Nagato Y. Aberrant Panicle Organization 1 temporally regulates meristem identity in rice. Dev Biol. 2005;282:349–60.

    CAS  PubMed  Google Scholar 

  23. Komatsu KK, Maekawa MM, Ujiie SS, Satake YY, Furutani II, Okamoto HH, et al. LAX and SPA: major regulators of shoot branching in rice. Proc Natl Acad Sci USA. 2003;100:11765–70.

    CAS  PubMed  PubMed Central  Google Scholar 

  24. Bai X, Huang Y, Mao D, Wen M, Zhang L, Xing Y. Regulatory role of FZP in the determination of panicle branching and spikelet formation in rice. Sci Rep. 2016;6:19022.

    CAS  PubMed  PubMed Central  Google Scholar 

  25. Bai X, Zhao H, Huang Y, Xie W, Han Z, Zhang B, et al. Genome-Wide Association Analysis Reveals Different Genetic Control in Panicle Architecture Between and Rice. The Plant Genome. 2016;9:0.

    Google Scholar 

  26. Zhang L, Yu H, Ma B, Liu G, Wang J, Wang J, et al. A natural tandem array alleviates epigenetic repression of IPA1 and leads to superior yielding rice. Nat Commun. 2017;8:14789.

    CAS  PubMed  PubMed Central  Google Scholar 

  27. Cubry P, Tranchant-Dubreuil C, Thuillet A-C, Monat C, Ndjiondjop M-N, Labadie K, et al. The rise and fall of african rice cultivation revealed by analysis of 246 new genomes. Curr Biol. 2018;28:2274-2282.e6.

    CAS  PubMed  Google Scholar 

  28. Jones MP, Dingkuhn M, Alukosnm GK, Semon M. Interspecific Oryza Sativa L. X O. Glaberrima Steud. progenies in upland rice improvement. Euphytica. 1997;94:237–46.

    Google Scholar 

  29. Linares OF. African rice (Oryza glaberrima): history and future potential. Proc Natl Acad Sci USA. 2002;99:16360–5.

    CAS  PubMed  PubMed Central  Google Scholar 

  30. Thiémélé D, Boisnard A, Ndjiondjop M-N, Chéron S, Séré Y, Aké S, et al. Identification of a second major resistance gene to Rice yellow mottle virus, RYMV2, in the African cultivated rice species O glaberrima. Theor Appl Genet. 2010;121:169–79.

    PubMed  Google Scholar 

  31. Wambugu PW, Ndjiondjop M-N, Henry R. Genetics and genomics of african rice (Oryza glaberrima Steud) domestication. Rice (N Y). 2021;14:6–14.

    PubMed  Google Scholar 

  32. Li X-M, Chao D-Y, Wu Y, Huang X, Chen K, Cui L-G, et al. Natural alleles of a proteasome α2 subunit gene contribute to thermotolerance and adaptation of African rice. Nat Genet. 2015;47:827–33.

    CAS  PubMed  Google Scholar 

  33. Lorieux M, A Garavito, J Bouniol, A Gutiérrez, M-N Ndjiondjop, R Guyo, CP Martinez, J Tohme, A Ghesquière. Unlocking the O. glaberrima treasure for rice breeding in Africa. In « Realizing Africa’s Rice Promise ». Marco Wopereis and David Johnson, Eds CABI, London.

  34. Harrop TWR, Mantegazza O, Luong AM, Béthune K, Lorieux M, Jouannic S, et al. A set of AP2-like genes is associated with inflorescence branching and architecture in domesticated rice. J Exp Bot. 2019;70:5617–29.

    CAS  PubMed  PubMed Central  Google Scholar 

  35. Cubry P, Pidon H, Ta KN, Tranchant-Dubreuil C, Thuillet A-C, Holzinger M, et al. genome wide association study pinpoints key agronomic QTLs in African rice Oryza glaberrima. Rice. 2020;13:66.

    PubMed  PubMed Central  Google Scholar 

  36. Eshed Y, Zamir D. An introgression line population of Lycopersicon pennellii in the cultivated tomato enables the identification and fine mapping of yield-associated QTL. Genetics. 1995;141:1147–62.

    CAS  PubMed  PubMed Central  Google Scholar 

  37. Koumproglou R, Wilkes TM, Townson P, Wang XY, Beynon J, Pooni HS, et al. STAIRS: a new genetic resource for functional genomic studies of Arabidopsis. Plant J. 2002;31:355–64.

    CAS  PubMed  Google Scholar 

  38. Balakrishnan D, Surapaneni M, Mesapogu S, Neelamraju S. Development and use of chromosome segment substitution lines as a genetic resource for crop improvement. Theor Appl Genet. 2019;132:1–25.

    CAS  PubMed  Google Scholar 

  39. Zhang Y, Zhou J, Xu P, Li J, Deng X, Deng W, et al. A Genetic resource for rice improvement: introgression library of agronomic traits for all AA genome oryza species. Front Plant Sci. 2022;13:856514.

    PubMed  PubMed Central  Google Scholar 

  40. Gutiérrez AG, Carabalí SJ, Giraldo OX, Martínez CP, Correa F, Prado G, et al. Identification of a Rice stripe necrosis virus resistance locus and yield component QTLs using Oryza sativa x O. glaberrima introgression lines. Bmc Plant Biol. 2010;10:6.

    PubMed  PubMed Central  Google Scholar 

  41. Shim RA, Angeles ER, Ashikari M, Takashi T. Development and evaluation of Oryza glaberrima Steud. chromosome segment substitution lines (CSSLs) in the background of O. sativa L. cv. Koshihikari. Breed Sci. 2010;60:613–9.

    Google Scholar 

  42. Crowell S, Falcão AX, Shah A, Wilson Z, Greenberg AJ, McCouch SR. High-resolution inflorescence phenotyping using a novel image-analysis pipeline. PANorama Plant Physiol. 2014;165:479–95.

    CAS  PubMed  Google Scholar 

  43. Crowell S, Korniliev P, Falcão A, Ismail A, Gregorio G, Mezey J, et al. Genome-wide association and high-resolution phenotyping link Oryza sativa panicle traits to numerous trait-specific QTL clusters. Nat Commun. 2016;7:10527.

    CAS  PubMed  PubMed Central  Google Scholar 

  44. Zhang Y, Ma Y, Chen Z, Zou J, Chen T, Li Q, Pan X, Zuo S. Genome-wide association studies reveal new genetic targets for five panicle traits of international rice varieties. Rice Sci. 2015;22:217–26.

    CAS  Google Scholar 

  45. Rebolledo MC, Peña AL, Duitama J, Cruz DF, Dingkuhn M, Grenier C, et al. Combining image analysis, genome wide association studies and different field trials to reveal stable genetic regions related to panicle architecture and the number of spikelets per panicle in rice. Front Plant Sci. 2016;7:1384.

    PubMed  PubMed Central  Google Scholar 

  46. Ta KN, Khong NG, HA TL, Nguyen DT, Mai DC, Hoang TG, Phung TPN, Bourrié I, Courtois B, Tran TTH, Dinh BY, Tuan N, Do NV, Lebrun M, Gantet P, Jouannic S. A genome-wide association study using a Vietnamese landrace panel of rice (Oryza sativa) reveals new QTLs controlling panicle morphological traits. BMC Plant Biol. 2018;18:282.

    PubMed  PubMed Central  Google Scholar 

  47. Zhang B, Shang L, Ruan B, Zhang A, Yang S, Jiang H, et al. Development of three sets of high-throughput genotyped rice chromosome segment substitution lines and QTL mapping for eleven traits. Rice. 2019;12:33.

    CAS  PubMed  PubMed Central  Google Scholar 

  48. Bai S, Hong J, Li L, Su S, Li Z, Wang W, et al. Dissection of the genetic basis of rice panicle architecture using a genome-wide association study. Rice. 2021;14:77.

    CAS  PubMed  PubMed Central  Google Scholar 

  49. Zhong H, Liu S, Meng X, Sun T, Deng Y, Kong W, et al. Uncovering the genetic mechanisms regulating panicle architecture in rice with GPWAS and GWAS. BMC Genomics. 2021;22:86.

    CAS  PubMed  PubMed Central  Google Scholar 

  50. Shao J, Haider I, Xiong L, Zhu X, Hussain RMF, Overnäs E, et al. Functional analysis of the HD-Zip transcription factor genes Oshox12 and Oshox14 in rice. PLoS ONE. 2018;13:e0199248.

    PubMed  PubMed Central  Google Scholar 

  51. Beretta VM, Franchini E, Din IU, Lacchini E, den Broeck LV, Sozzani R, et al. The ALOG family members OsG1L1 and OsG1L2 regulate inflorescence branching in rice. Plant J. 2023.

    Article  PubMed  Google Scholar 

  52. Masiero S, Imbriano C, Ravasio F, Favaro R, Pelucchi N, Gorla MS, et al. Ternary complex formation between MADS-box transcription factors and the histone fold protein NF-YB*. J Biol Chem. 2002;277:26429–35.

    CAS  PubMed  Google Scholar 

  53. Fornara F, Parenicová L, Falasca G, Pelucchi N, Masiero S, Ciannamea S, et al. Functional characterization of OsMADS18, a member of the AP1/SQUA subfamily of MADS box genes. Plant Physiol. 2004;135:2207–19.

    CAS  PubMed  PubMed Central  Google Scholar 

  54. Li F, Liu W, Tang J, Chen J, Tong H, Hu B, et al. Rice DENSE AND ERECT PANICLE 2 is essential for determining panicle outgrowth and elongation. Cell Res. 2010;20:838–49.

    PubMed  Google Scholar 

  55. Zhu Q-H, Hoque MS, Dennis ES, Upadhyaya NM. Ds tagging of BRANCHED FLORETLESS 1 (BFL1) that mediates the transition from spikelet to floret meristem in rice (Oryza sativa L). BMC Plant Biol. 2003;3:6.

    PubMed  PubMed Central  Google Scholar 

  56. Chen Q, Tian F, Cheng T, Jiang J, Zhu G, Gao Z, et al. Translational repression of FZP mediated by CU-rich element/OsPTB interactions modulates panicle development in rice. Plant J. 2022.

    Article  PubMed  PubMed Central  Google Scholar 

  57. Ali ML, Sanchez PL, Yu S, Lorieux M, Eizenga GC. Chromosome segment substitution lines: a powerful tool for the introgression of valuable genes from oryza wild species into cultivated rice (O. sativa). Rice. 2010;3:218–34.

    Google Scholar 

  58. Xing Y, Tan Y, Hua J, Sun X, Xu C, Zhang Q. Characterization of the main effects, epistatic effects and their environmental interactions of QTLs on the genetic basis of yield traits in rice. Theor Appl Genet. 2002;105:248–57.

    CAS  PubMed  Google Scholar 

  59. Jeon J-S, Jang S, Lee S, Nam J, Kim C, Lee S-H, et al. leafy hull sterile1 Is a homeotic mutation in a rice MADS Box gene affecting rice flower development. Plant Cell. 2000;12:871–84.

    CAS  PubMed  PubMed Central  Google Scholar 

  60. Prasad K, Parameswaran S, Vijayraghavan U. OsMADS1, a rice MADS-box factor, controls differentiation of specific cell types in the lemma and palea and is an early-acting regulator of inner floral organs. Plant J. 2005;43:915–28.

    CAS  PubMed  Google Scholar 

  61. Cui R, Han J, Zhao S, Su K, Wu F, Du X, et al. Functional conservation and diversification of class E floral homeotic genes in rice (Oryza sativa). Plant J. 2010;61:767–81.

    CAS  PubMed  Google Scholar 

  62. Khanday I, Yadav SR, Vijayraghavan U. Rice LHS1/OsMADS1 controls floret meristem specification by coordinated regulation of transcription factors and hormone signaling pathways. Plant Physiol. 2013;161:1970–83.

    CAS  PubMed  PubMed Central  Google Scholar 

  63. Hill MS, Zande PV, Wittkopp PJ. Molecular and evolutionary processes generating variation in gene expression. Nat Rev Genet. 2021;22:203–15.

    CAS  PubMed  Google Scholar 

  64. Wray GA. The evolutionary significance of cis-regulatory mutations. Nat Rev Genet. 2007;8:206–16.

    CAS  PubMed  Google Scholar 

  65. Mao C, He J, Liu L, Deng Q, Yao X, Liu C, et al. OsNAC2 integrates auxin and cytokinin pathways to modulate rice root development. Plant Biotechnol J. 2020;18:429–42.

    CAS  PubMed  Google Scholar 

  66. Wang J, Bao J, Zhou B, Li M, Li X, Jin J. The osa-miR164 target OsCUC1 functions redundantly with OsCUC3 in controlling rice meristem/organ boundary specification. New Phytol. 2021;229:1566–81.

    CAS  PubMed  Google Scholar 

  67. Zhao Y, Hu Y, Dai M, Huang L, Zhou D-X. The WUSCHEL-related homeobox gene WOX11 is required to activate shoot-borne crown root development in rice. Plant Cell. 2009;21:736–48.

    CAS  PubMed  PubMed Central  Google Scholar 

  68. Zhao Y, Cheng S, Song Y, Huang Y, Zhou S, Liu X, et al. The interaction between rice ERF3 and WOX11 promotes crown root development by regulating gene expression involved in cytokinin signaling. Plant Cell Online. 2015;27:2469–83.

    CAS  Google Scholar 

  69. Cheng S, Huang Y, Zhu N, Zhao Y. The rice WUSCHEL-related homeobox genes are involved in reproductive organ development, hormone signaling and abiotic stress response. Gene. 2014;549:266–74.

    CAS  PubMed  Google Scholar 

  70. Cheng S, Zhou D-X, Zhao Y. WUSCHEL-related homeobox gene WOX11 increases rice drought resistance by controlling root hair formation and root system development. Plant Signal Behav. 2015;11:e1130198.

    PubMed Central  Google Scholar 

  71. Zhou S, Jiang W, Long F, Cheng S, Yang W, Zhao Y, et al. Rice homeodomain protein WOX11 recruits a histone acetyltransferase complex to establish programs of cell proliferation of crown root meristem. Plant Cell. 2017;29:1088–104.

    CAS  PubMed  PubMed Central  Google Scholar 

  72. Zhang T, Li R, Xing J, Yan L, Wang R, Zhao Y. The YUCCA-Auxin-WOX11 module controls crown root development in rice. Front Plant Sci. 2018;9:523.

    PubMed  PubMed Central  Google Scholar 

  73. Cheng S, Tan F, Lu Y, Liu X, Li T, Yuan W, et al. WOX11 recruits a histone H3K27me3 demethylase to promote gene expression during shoot development in rice. Nucleic Acids Res. 2018;46:gky017-.

    Google Scholar 

  74. Ren D, Li Y, Zhao F, Sang X, Shi J, Wang N, et al. MULTI-FLORET SPIKELET1, which encodes an AP2/ERF protein, determines spikelet meristem fate and sterile lemma identity in rice. Plant Physiol. 2013;162:872–84.

    CAS  PubMed  PubMed Central  Google Scholar 

  75. Wang S-S, Chung C-L, Chen K-Y, Chen R-K. A novel variation in the FRIZZLE PANICLE (FZP) gene promoter improves grain number and yield in rice. Genetics. 2020. 

    Article  PubMed  PubMed Central  Google Scholar 

  76. Al-Tam FM, Adam H, Anjos AD, Lorieux M, Larmande P, Ghesquière A, et al. P-TRAP: a panicle trait phenotyping tool. BMC Plant Biol. 2013;13:122.

    PubMed Central  Google Scholar 

  77. Serret J, Mariac C, Albar L, Sabot F. From low cost plant HMW DNA extraction to MinION sequencing v1. 2021.

  78. Kolmogorov M, Yuan J, Lin Y, Pevzner PA. Assembly of long, error-prone reads using repeat graphs. Nat Biotechnol. 2019;37:540–6.

    CAS  PubMed  Google Scholar 

  79. Alonge M, Lebeigle L, Kirsche M, Jenike K, Ou S, Aganezov S, et al. Automated assembly scaffolding using RagTag elevates a new tomato system for high-throughput genome editing. Genome Biol. 2022;23:258.

    CAS  PubMed  PubMed Central  Google Scholar 

  80. Manni M, Berkeley MR, Seppey M, Zdobnov EM. BUSCO: assessing genomic data quality and beyond. Curr Protoc. 2021;1:e323.

    PubMed  Google Scholar 

  81. Tranchat-Dubreuil C, Chenal C, Blaison M, Laurence Albar L, Klein V, Mariac C, Wing R, Vigouroux Y, Sabot F. FrangiPANe, a tool for creating a panreference using left behind reads. Nar Genom Bioinform. 2023;5:lqad013.

    Google Scholar 

  82. Shen W, Le S, Li Y, Hu F. SeqKit: a cross-platform and ultrafast toolkit for FASTA/Q file manipulation. PLoS ONE. 2016;11:e0163962.

    PubMed  PubMed Central  Google Scholar 

  83. Huang F, Jiang Y, Chen T, Li H, Fu M, Wang Y, et al. New data and new features of the funricegenes (Functionally Characterized Rice Genes) database: 2021 update. Rice. 2022;15:23.

    PubMed  PubMed Central  Google Scholar 

  84. Harrop TWR, Din IU, Gregis V, Osnato M, Jouannic S, Adam H, et al. Gene expression profiling of reproductive meristem types in early rice inflorescences by laser microdissection. Plant J. 2016;86:75–88.

    CAS  PubMed  Google Scholar 

  85. Thompson JD, Higgins DG, Gibson TJ. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 1994;22:4673–80.

    CAS  PubMed  PubMed Central  Google Scholar 

  86. Chow C-N, Lee T-Y, Hung Y-C, Li G-Z, Tseng K-C, Liu Y-H, et al. a new and updated resource for reconstructing transcriptional regulatory networks from ChIP-seq experiments in plants. Nucleic Acids Res. 2018;47 Database issue:gky1081-.

    Google Scholar 

  87. Gel B, Serra E. karyoploteR: an R/Bioconductor package to plot customizable genomes displaying arbitrary data. Bioinformatics. 2017;33:3088–90.

    CAS  PubMed  PubMed Central  Google Scholar 

Download references


The authors acknowledge Ndomassi Tando and the ISO 9001 certified IRD iTrop HPC (member of the South Green Platform) at IRD Montpellier for providing HPC resources that have contributed to the research results reported within this paper (URL: We thank Lady Johanna Arbelaez Rivera (Alliance Bioversity-CIAT), Sophie Chéron (IRD) and Harold Chrestin (IRD) for plant care. We thank Marie-Christine Combes (IRD) and Carole Gauron (IRD) for their help with the panicle phenotyping work. We thank Jean-Francois Rami (CIRAD) for his help with R scripts. We also thank Laurence Albar (IRD) for her helpful feedback on the manuscript.


This research was funded by the Agropolis Foundation through the “Investissements d'avenir” programme (ANR-10-LABX-0001–01), the Fondazione Cariplo (EVOREPRICE 1201–004), and the CGIAR Research program on Rice (RICE CRP).

Author information

Authors and Affiliations



HA, ML designed the project, with input from SJ; HA and ML performed QTL analysis; ML designed the population subset and the markers and supervised field phenotyping; AG performed the SSR analysis on the BC3DH and BC4 plants. MC and JS performed the ONT sequencing. MC and FS performed genome assembly and annotation transfer. FN performed the bioinformatic minimap2 alignments. JO developed a Python script for the colocalization analysis. HA, SJ and ML analyzed and interpreted the data. HA wrote the initial draft of the paper. SJ, JT and ML revised the paper. All authors critically reviewed and approved the final manuscript.

Corresponding authors

Correspondence to Hélène Adam, Stefan Jouannic or Mathias Lorieux.

Ethics declarations

Ethics approval and consent to participate

Experimental research and field studies on plants (either cultivated or wild), including the collection of plant material, complies with relevant institutional, national, and international guidelines and legislation.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1:

Table S1. CSSL information showing Og_MG12 introgression(s) for each line and their localization(s) (chromosome and left-right SSR markers). Target fragments were those selected so as to constitute the whole Og_MG12 genome with the 60 CSSLs. Table S2. Quantification of panicle traits in the CSSLs and parents Os_Caiapó and O.g_MG12. Abbreviations: RL, Rachis Length; PBN, Primary Branch Number; PBL, Primary Branch Length; SBN, Secondary Branch Number; SpN, Spikelet Number; SBN/PBN, ratio of SBN and PBN. Table S3. Variance analysis done on panicle traits for the 60 CSSLs and their parents phenotyped the year 1. Table S4. Significant line*trait associations observed in the population studied based on a Dunnett's test. Table S5. QTL and GWAS sites colocalized with the QTLs detected in this study. Table S6. Metrics of genome quality obtained for Os_Caiapó and Og_MG12 genome ONT sequencing assemblies. Table S7. Summary of the positions (in bp) of structural variations observed in the QTLs detected. Table S8. Genes described in the FunRiceGenes database located within each QTL detected in this study. Table S9. Gene synteny and description of loci annotated between the RM10 and RM420 markers in Os_Caiapó (RAP_db and MSU annotations) and Og_MG12 (RAP_db, MSU and OglaRS2). Genes highlighted in yellow are absent in the q_7 region of Og_MG12, genes highlighted in green are annotated in the Og_MG12 q_7 region but are absent in the Os_Caiapó genome, genes highlighted in light green are duplicated in the Og_MG12 q_7 region and genes highlighted in pink are annotated in the Og_MG12 q_7 region, are absent in the Os_ Caiapó q_7 region but are present in other chromosomes in the Os_Caiapó genome. Table S10. Amino acid modifications in proteins encoded by candidate genes in q_7. Table S11. TFBS variations in the promoters of candidate genes present in q_7. Abbreviations: EIN3, ETHYLENE-INSENSITIVE-LIKE3; C2H2, C2H2-Type Zinc finger; AP2, AP2-ERF; bHLH, basic helix-loop-helix; TCP, TEOSINTE-BRANCHED1; SBN, SQUAMOSA BINDING PROTEIN; NY-YB, nuclear factor Y; HB, HOMEODOMAIN; MADS, MADS-BOX; bZIP, basic leucine zipper; TBP, TATA-box-binding.

Additional file 2:

Figure S1. CSSL Finder screenshots showing graphical representations of the genotypes of the 60 BC3DH lines along with corresponding data for each evaluated panicle trait across the two repetitions together and for each repetition separated (rep1 and rep2). (a) RL, Rachis Length; (b) SBN/PBN, Secondary Branch Number per Primary Branch; (c) SpN. Spikelet Number; (d) PBN, Primary Branch Number; (e) PBL, Primary Branch Length; (f) TBN, Tertiary Branch Number.. The 12 chromosomes are displayed vertically. They are covered by 200 evenly dispersed SSR markers. The genotypes of the individual lines are displayed horizontally. Shading indicates the allelic composition of chromosomes. Light gray areas represent the Os_Caiapó genetic background, black areas represent the Og_MG12 chromosome segments, dark gray areas represent the heterozygous segments and blue areas correspond to missing data. On the right, solid color bars indicate the values of the panicle traits tested for each line. At the bottom of each graph, the dotted line indicates the statistical threshold of the F-test for the evaluated panicle trait. Figure S2. Graphic representation of the genotypes of the 60 BC3DH CSSLs, showing line x trait significant associations and QTL positions. The 12 chromosomes are covered by 200 evenly dispersed SSR markers. Genotypes are displayed horizontally. Black areas represent the O. glaberrima MG12 target chromosome segments, the set of segments broadly covering the entire O. glaberrima MG12 genome. White areas represent the O. sativa Caiapó genetic background and grey areas represent O. glaberrima additional chromosomal segments. Phenotypic effects of CSSL lines are represented in relative terms as circles on the right of the figure. The area of each circle is proportional to its relative effect. Horizontal bars at the bottom of the figure indicate QTL positions deduced from analyses performed using CSSL finder software. The colors of the circles and bars represent the effect as follows: red, increasing effect compared to the recurrent parent Os_Caiapó; green; decreasing effect. Figure S3. Genetic and phenotypic description of CSSLs showing the extreme significant differences for the panicle trait analyzed (A) SBN/PBN, (B) SpN, (C) PBN, (D) PBL and (E) TBN. Upper subpanel: Chromosome graphical representation of Og_MG12 introgression positions in the CSSL. Position of the SSR markers is indicated in Mbp. Lower subpanel: boxplots of the phenotypic variation observed in Year 1 and Year 2 in these lines. Each point represents the phenotypic value for one panicle. Statistical significance (t-test p-values) between Os_Caiapó and each line for the panicle morphological trait is indicated as follows: NS if the test is non-significant; *p-values < 0.05; **<0.01; ***<0.001. Abbreviations: PBN, primary branch number; PBL, primary branch length; TBN, tertiary branch number; SBN/PBN, ratio between secondary branch and primary branch numbers; SpN, Spikelet number. Figure S4. Phenotypic description of BC4 lines to dissect the effects of QTLs on panicle traits. Boxplot of the phenotypic variation observed in Os_Caiapó (yellow), Og_MG12 (green), and BC4 (blue) lines in greenhouse (2021). Each point represents the phenotypic value for one panicle. Statistical significance (t-test p-values) between Os_Caiapó and each line for the panicle morphological traits is indicated as follows: ** p-values <0.01; ***<0.001. Abbreviations: PBN, primary branch number; PBL, primary branch length; TBN, tertiary branch number; SBN/PBN, ratio between secondary branch and primary branch numbers; SpN, spikelet number. Figure S5. Dotplots of the minimap2 alignment (implemented in D-GENIES web facilities) of panicle trait-related QTLs detected between the Os_Caiapó and Og_MG12 genomes and between each of the latter aligned against O. sativa cv. Nipponbare as a reference genome. QTL coordinates in each genome are indicated on axis X and Y. Dot colors are relative to the identity value (I) which is a BLAST-like alignment identity (I=Number of bases, including gaps per number of matching bases in the bases). Figure S6. Physical map positions of detected QTLs and colocalizations with known genes related to branching and flowering.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit The Creative Commons Public Domain Dedication waiver ( applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Adam, H., Gutiérrez, A., Couderc, M. et al. Genomic introgressions from African rice (Oryza glaberrima) in Asian rice (O. sativa) lead to the identification of key QTLs for panicle architecture. BMC Genomics 24, 587 (2023).

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: