Genetic and gene expression analysis of flowering time regulation by light quality in lentil

Abstract

Background and Aims

Flowering time is important due to its roles in plant adaptation to different environments and subsequent formation of crop yield. Changes in light quality affect a range of developmental processes including flowering time, but little is known about light quality-induced flowering time control in lentil. This study aims to investigate the genetic basis for differences in flowering response to light quality in lentil.

Methods

We explored variation in flowering time caused by changes in red/far-red-related light quality environments of a lentil interspecific recombinant inbred line (RIL) population developed from a cross between Lens culinaris cv. Lupa and L. orientalis accession BGE 016880. A genetic linkage map was constructed and then used for identifying quantitative trait loci (QTLs) associated with flowering time regulation under different light quality environments. Differential gene expression analysis through transcriptomic study and RT-qPCR were used to identify potential candidate genes.

Key Results

QTL mapping located 13 QTLs controlling flower time under different light quality environments, with phenotypic variance explained ranging from 1.7 to 62.9 %. Transcriptomic profiling and gene expression analysis for both parents of this interspecific RIL population identified flowering-related genes showing environment–specific differential expression (flowering DEGs). One of these, a member of the florigen gene family FTa1 (LcFTa1), was located close to three major QTLs. Furthermore, gene expression results suggested that two other florigen genes (LcFTb1 and LcFTb2), MADS-box transcription factors such as LcAGL6/13d, LcSVPb, LcSOC1b and LcFULb, as well as bHLH transcription factor LcPIF6 and Gibberellin 20 oxidase LcGA20oxC,G may also be involved in the light quality response.

Conclusions

Our results show that a major component of flowering time sensitivity to light quality is tightly linked to LcFTa1 and associated with changes in its expression. This work provides a foundation for crop improvement of lentil with better adaptation to variable light environments.

INTRODUCTION

Flowering, a transition from vegetative growth to reproductive growth, is one of the most important events for flowering plants. The switch from vegetative to reproductive growth is essential for crop production when seeds or fruits are to be the end products. Plants need to recognize and process a wide range of environmental and internal cues, and then consolidate these into a single effective developmental choice: to flower or not (Putterill et al., 2004). Some of the main environmental factors that control flowering time include photoperiod and temperature, but light quality is also known to have an important influence (Adams et al., 2009; Casal, 2013), and acts as a crucial signal that communicates the close presence of neighbouring plants (Adams et al., 2009). One of the main components of this signal is a reduction in the red to far-red ratio (R/FR), which results from the selective absorption of red wavelengths by chlorophyll (Smith and Whitelam, 1997). This reduction in R/FR is primarily sensed through the phytochrome family of photoreceptors, which mediate its effects on a wide range of developmental processes, including increased stem elongation, reduced branching and accelerated flowering (Smith and Whitelam, 1997; Quail, 2002; Kami et al., 2010). These responses are known as shade avoidance syndrome (SAS) which usually leads to the re-allocation of resources with a concomitant effect on plant fitness, flowering time and yield (Page et al., 2011; Lockhart, 2014; Tang and Liesche, 2017; Rea, 2019). In a crop situation, SAS is invoked when plants are growing close enough together to alter the light quality. This has an impact on optimal seeding rates, critical weed thresholds and understanding how crops work together in an intercropping system. We can simulate these conditions in a controlled manner by changing lighting in a growth chamber to manipulate R/FR. Knowledge of molecular mechanisms of shade avoidance will help us to better understand SAS and breed for varieties that respond in a more positive manner (Kebrom and Brutnell, 2007; Merotto et al., 2009; Tang and Liesche, 2017).

Studies of flowering time control have identified numerous flowering time genes and defined genetic regulatory networks in model species (Song et al., 2015; Hori et al., 2016; Gol et al., 2017), as well as in some legumes (Weller and Ortega, 2015; Ridge et al., 2017; Ortega et al., 2019; Lin et al., 2021). Many of these genes contribute to the natural genetic variation for flowering time and for related growth and yield traits, and provide adaptation of crop plants to various locations and management practices (Jung and Müller, 2009; Bouchet et al., 2013; Blümel et al., 2015). The importance and genetic basis of variation in response to light quality specifically is relatively poorly understood in many crop species, although many of the genes defined in Arabidopsis are widely conserved (Hecht et al., 2005; Adams et al., 2009; Leijten et al., 2018).

A genetic understanding of flowering time regulation is an important general objective in plant breeding, as it facilitates the generation of new varieties that will be better adapted to a specific environment. Genetic analysis of flowering time has been conducted in many legume crops (Pérez-Vega et al., 2010; Cruz-Izquierdo et al., 2012; Upadhyaya et al., 2015; Liu et al., 2016; Ridge et al., 2017), including lentil (Sarker et al., 1999; Fratini et al., 2007; Tullu et al., 2008; Fedoruk et al., 2013). However, few studies so far have investigated the genetic basis for flowering responses to light quality, and none in lentil.

Cultivated lentil (Lens culinaris Medik.) is the third most important cool-season grain legume (FAO, 2015). Lentils have high nutritional and health benefits, and like other crop legumes, play a significant role in supporting environmentally sustainable agriculture due to their nitrogen fixation ability. In addition to the cultivated lentil, there are six wild species within the genus Lens and L. orientalis is generally considered to be the wild progenitor (Mayer and Soltis, 1994; Van Oss et al., 1997). Rich genetic diversity identified in wild Lens species makes them good candidates for the genetic improvement of cultivated lentils (Vail et al., 2011; Podder et al., 2013; Bhadauria et al., 2017). In a recent study, we found that the flowering time of most wild lentil genotypes was not significantly affected by light quality changes, whereas it was consistently accelerated under low R/FR conditions in cultivated lentil (Yuan et al., 2017). This variation in flowering time sensitivity toward light quality change indicated that genes or specific alleles associated with this trait could be used to select or modify flowering time in cultivated lentil.

In this study, we used an interspecific recombinant inbred line (RIL) population of lentil developed by crossing a cultivated lentil with an L. orientalis accession (Fratini et al., 2007) where the parents had contrasting flowering responses to light quality change (Yuan et al. 2017). We characterized the variations in flowering responses of individual RILs to contrasting R/FR environments and used a high-density genetic linkage map to identify quantitative trait loci (QTLs) associated with differential sensitivity to light quality. De novo transcriptomic analysis was used to characterize the effects of light quality on gene expression, and together with the newly available L. culinaris genome (CDC Redberry, v 2.0) (Ramsay et al., 2021), hereby referred to as the reference genome, we were able to identify and evaluate potential candidate genes for this important response.

MATERIALS AND METHODS

Plant material, growth conditions and phenotypic evaluation

An interspecific RIL population developed from a cross between L. culinaris cv. Lupa and L. orientalis accession BGE 016880 (Fratini et al., 2007) was used in this study. Both parents have contrasting flowering time responses to light quality change, with the wild parent being less sensitive (Yuan et al., 2017). The RIL population of 93 individuals and parents was grown and evaluated in two Conviron GR178 walk-in plant growth chambers with contrasting R/FR ratios, but similar light quantities based on the photosynthetically active radiation (PAR). The high R/FR light quality condition was the typical state in the growth chamber fitted with T5 835 High Output fluorescent bulbs (Philips, Andover, MA, USA). The low R/FR light quality condition was reached by adding evenly spaced PfrSpec™ LED light panels (model RAY44, peak spectrum at 730 nm; Fluence Bioengineering, Inc., Austin, TX, USA) into the light bank fitted with the T5 835 High Output fluorescent bulbs. The spectral photon flux and PAR of each light condition was measured using a spectroradiometer (Model PS-300; Apogee Instruments, Logan, UT, USA). The R and FR values were calculated using the spectral photon flux at 650–670 and 720–740 nm, respectively (Smith, 1982). The high R/FR condition had an R/FR ratio of 7.30 ± 0.14 with PAR at 402.2 ± 33.6 µmol m^–2 s^–1, and the low R/FR condition had an R/FR ratio of 0.19 ± 0.01 with a PAR level very close to the high R/FR condition at 395.6 ± 32.9 µmol m^–2 s^–1. The spectral distribution of these two light conditions is shown in Fig. 1.

Fig. 1.

Spectral distribution of the light quality treatments used to compare the effect of light quality on days to flower of a lentil interspecific recombinant inbred line (RIL) population developed from L. culinaris cv. Lupa and L. orientalis BGE 016880.

Prior to planting, seeds of all RILs and parents were stored at −20 °C for 1 week and then the seed coats were nicked to improve imbibition and germination. Square, 10-cm pots were filled with growth medium consisting of 50 % Sunshine Mix #3 and 50 % Sunshine Mix #4 (Sun Gro Horticulture Canada Ltd, Seba Beach, AB, Canada). Two seeds were sown in each pot and thinned to a single plant after emergence. The RILs and the parents were planted in a completely randomized design with four technical replicates per genotype under each light quality condition and the experiment was repeated once. Plants were grown and maintained at 22 °C/16 h day and 16 °C/8 h night for both conditions and repeats. Days to flower (DTF) were calculated based on days from emergence to R1 (one open flower at any node) (Erskine et al., 1990). Levene’s test of homogeneity of variance was done for days to flower using R (R Core Team, 2013). There were no significant effects of repeat, so data were averaged over the two repeats. For each individual, the trait ‘Flowering Time Sensitivity’ (FTS) was calculated as the ratio of the difference in DTF between two conditions divided by the sum of the DTF under two conditions to avoid bias due to the underlying differences in DTF.

$${\displaystyle \begin{array}{l}{\displaystyle \mathrm{F}\mathrm{T}\mathrm{S}=(\mathrm{D}\mathrm{T}{\mathrm{F}}_{\mathrm{h}\mathrm{i}\mathrm{g}\mathrm{h}}-\mathrm{D}\mathrm{T}{\mathrm{F}}_{\mathrm{l}\mathrm{o}\mathrm{w}})/(\mathrm{D}\mathrm{T}{\mathrm{F}}_{\mathrm{h}\mathrm{i}\mathrm{g}\mathrm{h}}+\mathrm{D}\mathrm{T}{\mathrm{F}}_{\mathrm{l}\mathrm{o}\mathrm{w}}).}\end{array}}$$

RADseq library preparation, sequencing and SNP calling

Illumina sequencing libraries were prepared using genomic DNA from the RIL population for restriction site-associated DNA sequencing (RADseq). HindIII and NlaIII were chosen as two restriction enzymes to digest the genomic DNA. Detailed library preparation procedures were described in von Wettberg et al. (2018). Fragments were sequenced as single-end 100-bp reads on an Illumina HiSeq4000 at the University of California at Davis Genome Core Facility. Reads were mapped to the L. culinaris v2.0 reference genome using Bowtie allowing only end-to-end matches (Langmead and Salzberg, 2012). Variant calling was performed with Samtools (Li et al., 2009) and output in VCF format (Danecek et al., 2011). Single nucleotide polymorphism (SNP) variants were filtered using VCFtools (Danecek et al., 2011) to exclude sites that contain an indel, keep variants that have no more than 66 % missing data with a minimum quality score of 30, and include only bi-allelic sites. After filtering, SNP variants were converted to ABH format, and markers that were 100 % identical in sequential order along the chromosome (including missing) were binned and further filtered for markers where only the parents were different. Initially, 84 721 SNPs were generated from the RIL population and, after filtering and binning, 15 686 SNPs were retained for linkage mapping.

Linkage map construction and QTL mapping

Linkage map construction was performed with IciMapping software (Meng et al., 2015). BIN functionality within the software was used to remove redundant SNP markers that have identical segregation in the RIL population. After binning, SNP markers were grouped using a logarithm of the odds (LOD) threshold value of 8.0. Linkage groups (LGs) were assigned using the genomic position of SNP markers on the L. culinaris v2.0 reference genome (Ramsay et al., 2021). The REcombination Counting and ORDering (RECORD) algorithm (Van Os et al., 2005) was used to order the SNPs within each LG. The Kosambi mapping function (Kosambi, 1943) was used to convert the recombination fractions into additive genetic distance (centiMorgans, cM). Rippling of the SNP markers was performed by permutation of a window of eight markers and using sum of adjacent recombination frequencies (SARF) as the rippling criterion. The linkage map was further adjusted using the R/qtl package (Broman et al., 2003). The marker order within an LG that gave the shortest genetic distance was chosen as the final map. Visualization of the linkage map was generated using the LinkageMapView package in R (Ouellette et al., 2018). ClicO was used to generate the Circos circular plot (Cheong et al., 2015) and the R/qtl package was used to estimate recombination fractions and LOD scores for all pairs of markers and generate the heatmap plot (Broman et al., 2003).

QTL mapping was performed using BIP functionality in IciMapping software (Meng et al., 2015). For QTL identification, the genotyping data were integrated with the phenotypic data. Inclusive Composite Interval Mapping of Additive (ICIM-ADD) was used to detect additive QTLs. The walking speed chosen for QTL mapping was 1 cM. A 1000 permutation test was applied to decide the LOD thresholds (P ≤ 0.05) to determine the significance of identified QTLs. A one-LOD support interval was calculated for each QTL to obtain a 95 % confidence interval. The percentage of phenotypic variance explained by each QTL in proportion to the total phenotypic variance was estimated and QTLs with a positive or negative additive effect for the trait imply that the increase in the phenotypic value of the trait is contributed by alleles from L. culinaris cv. Lupa or L. orientalis BGE 016880, respectively. QTLs for all traits were named according to McCouch et al. (1997) and alphabetical order was used for QTLs on the same linkage group.

Transcriptomic analysis and RT-qPCR verification of the candidate genes

Both L. culinaris cv. Lupa and L. orientalis BGE 016880 were grown under two different R/FR environments, the same as those used for RIL population screening. Leaf sample collections started 2 weeks after emergence (T1 stage) and continued once a week for 4 more weeks (T2 to T5 stage). The T1 sampling stage was 2 weeks before L. orientalis BGE 016880 flowered under low R/FR environment, and Supplementary Data File S1 – Fig. S2 shows relationships between the sampling stages and the flowering times for both BGE 016880 and Lupa under different R/FR environments. Samples were taken at the same time of the day for each collection. Each biological replicate consisted of leaf material collected from three individual plants and three biological replicates were used.

Total RNA was isolated using an RNeasy Plant Mini Kit (Qiagen, Germantown, MD, USA) according to the protocol provided with the kit. On-column DNase digestion was performed during the isolation process according to the kit instructions. Extracted RNA was quantified and qualified using a NanoDrop 8000 UV–Vis spectrophotometer (NanoDrop, Wilmington, DE, USA) and an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) using an Agilent RNA 6000 Nano Assay. RNA sequencing (RNAseq) libraries were constructed using an Illumina TruSeq Stranded mRNA Sample Preparation Kit (Illumina Inc., San Diego, CA, USA) according to the protocol, and pooled libraries of 20 barcoded samples were sent for pair-end sequencing using an Illumina HiSeqTM 2500 sequencing platform (Illumina Inc., San Diego, CA, USA).

Qualities of the raw reads were checked using FastQC (Andrews, 2010) and adaptor sequences were removed using Trimmomatic (Bolger et al., 2014). Trimmed reads were used for de novo assembly of transcriptomes using Trinity (Grabherr et al., 2011). Salmon (Patro et al., 2017) was used to quantify the abundance of transcripts and 3D RNA-seq pipeline (Calixto et al., 2018; Guo et al., 2019) was used for the analysis of differential gene expression within species. Gene expression under the high R/FR environment was used as the control to analyse gene expression changes under the low R/FR environment for both BGE 016880 and Lupa. A detailed method regarding differential gene expression analysis is described in Supplementary Data Methods S1. Genes with false discovery rate (FDR)-adjusted P-values < 0.05 and |log₂FC|≥1 were considered differentially expressed genes (DEGs; FC, fold change).

A list of 244 lentil genes that showed high homology with genes involved in flowering in Arabidopsis thaliana (https://www.mpipz.mpg.de/14637/Arabidopsis_flowering_genes; and Higgins et al., 2010) and Medicago truncatula (Hecht et al., 2005; Putterill et al., 2013) was curated (Supplementary File S2) using in-house BLASTn search (https://knowpulse.usask.ca/blast/nucleotide/nucleotide) against the L. culinaris cv. CDC Redberry v2.0 reference genome. An e-value cut-off of 1e-5 or matched gene annotation in the lentil genome assembly was used to select the putative homologue. This list was used to explore the expression of known flowering genes in the lentil genome.

To confirm the reliability of RNAseq results, representative flowering-related genes were chosen for RT-qPCR using FAST SYBR® Green qPCR Master Mix (Applied Biosystems, Foster City, CA, USA) with a BioRad Real-Time PCR system. Actin (Lcu.2RBY.L011470) was selected as the reference gene. To derive the relative expression value, the delta-delta CT method was adopted (Livak and Schmittgen, 2001) and samples from the T1 stage in the low R/FR light quality environment were used as reference samples. The sequences of the primers used in qPCR are listed in Supplementary Data File S1 – Table S1.

Data availability

The RADseq data of the RIL population plus two parents (L. culinaris cv. Lupa and L. orientalis BGE 016880) as well as the raw RNAseq data from L. culinaris cv. Lupa and L. orientalis BGE 016880 have been submitted to the NCBI Sequence Read Archive with IDs PRJNA693555 and PRJNA693582. Related tables, figures and methods are given in Supplementary Data File S1. All other related data as well as the in-house developed scripts are given in Supplementary Data Files S2–S6.

RESULTS

Variation in flowering time response to light quality

To explore the genetic control of differences in flowering response to light quality that we had previously described for wild and domesticated lentil accessions (Yuan et al., 2017), we used an RIL population developed from an interspecific cross between L. culinaris cv. Lupa and an accession (BGE 016880) of L. orientalis, the putative progenitor of cultivated lentil (Mayer and Soltis, 1994). Flowering time in this population was assessed under two light environments with equivalent PAR but differing in R/FR (Fig. 1). We also expressed the response of each line to light quality as a ‘flowering time sensitivity’ index (FTS), which was calculated as (DTF_high − DTF_low)/(DTF_high + DTF_low), in order to avoid bias due to the underlying differences in DTF.

The wild parent, BGE 016880, flowered significantly earlier than the cultivated parent, Lupa, under both light quality environments and the wild parent had a low FTS compared to the cultivated parent (P < 0.05; Fig. 2 and Supplementary Data File S1 – Fig. S1). The RIL population segregated for days to flowering and for sensitivity to change in light quality, with probability density of the distributions centred on the mid-parent value under both light conditions and transgressing the parental range in both directions (Fig. 2). Some indication of a bimodal distribution was evident for flowering under high R/FR and for flowering time sensitivity (Fig. 2), suggesting the possibility of a substantial contribution from a single major locus.

Fig. 2.

Days to flower (DTF) of individuals from a lentil interspecific RIL population developed from a cross between L. culinaris cv. Lupa and L. orientalis BGE 016880, grown under light quality environments differing in red to far-red ratio (R/FR) (left and middle), and flowering time sensitivity (FTS) to changes in R/FR light quality environment (right). The violin plot illustrates the kernel probability density and the width of the shaded area represents the proportion of the data located there. The inner section of the violin plot shows the box plot indicating the median and interquartile range, and the 95 % confidence interval is shown by the whiskers. Dots outside the boxplot represent datapoints that are more than 1.5 times the upper quartile. The green arrow points to where the wild parent L. orientalis BGE 016880 falls in the plot and the brown arrow to the cultivated parent L. culinaris cv. Lupa.

Linkage map construction

To support QTL analysis of the observed differences, we generated SNP genotyping profiles for the Lupa × BGE 016880 population through a RADseq approach and used these to generate a genetic linkage map. A total of 4073 SNP markers were mapped into six LGs (Fig. 3A and Supplementary Data File S3). These groups were designated LG1–LG6 in accordance with their relationship to chromosomes of the reference genome. Markers from chromosome 7 mapped with those from chromosome 2 in LG2, a pseudo-linkage most likely due to a chromosomal rearrangement in one parent relative to the other. Rearrangements have previously been reported in Lens interspecies crosses (Ladizinsky, 1979) and are also apparent in our recent genome assembly efforts across Lens species (Ramsay et al., 2021). Respective marker positions between the linkage map and the reference genome are shown in Fig. 3B. A heat map of pairwise recombination fractions and LOD scores between all pairs of markers was visualized to check the quality of the linkage map. The tight linkage of the markers within each LG indicated the reliability of the developed linkage map (Fig. 3C). The total length of the linkage map spanned 5923.3 cM with LG2 being the largest, due to pseudolinkage of two chromosomes, and LG1 the smallest. The map distance is probably expanded due to unusual recombination, the mapping function or genotyping errors (Sybenga, 1996; Broman and Weber, 2000). Lentil chromosomes are quite large (400–600 Mb) (Ramsay et al., 2021), which could result in double cross-overs in a linkage map that would look like genotyping errors. We chose to leave them in the map in case they were real and accept the fact that they lead to expanded LG sizes. The number of markers per LG varied from 496 in LG1 to 958 in LG2, with an average distance between two adjacent markers of 1.5 cM and the largest gap at 9.4 cM in LG5 (Supplementary Data File S1 – Table S2).

Fig. 3.

(A) Genetic linkage map of a lentil interspecific RIL population developed from a cross between L. culinaris cv. Lupa and L. orientalis BGE 016880. Scale bar at the left of the linkage map is in centimorgans (cM). Bottom colour scheme shows the density of markers with a sliding window of 30 markers for calculation, and horizontal lines in each linkage group indicate the position of markers. (B) Circular representation of the markers on the genetic linkage map of the RIL population and their respective positions on the Lens culinaris reference genome. The scale on the outer ring for genetic linkage maps (LG1–LG6) is in cM and for the reference genome (LcuChr1–LcuChr7) it is in million base pairs (Mbp). (C) Validation of the map using pairwise linkage information. The plot shows the estimated recombination fractions and LOD scores for all pairs of markers after ordering the markers on each linkage group. The recombination fractions are in the upper left triangle while the LOD scores are in the lower right triangle. Estimates are plotted as a heat map with dark blue signifying no linkage and yellow representing tight linkage with low RF and large LOD. The diagonal yellow line indicates good linkage within each LG. LG2 is the one that is complicated by pseudolinkage of markers from two different L. culinaris chromosomes.

Genetic analysis of flowering time and sensitivity to R/FR

We detected six QTLs for flowering time under high R/FR light conditions (qDTFH-1A, qDTFH-1B, qDTFH-3, qDTFH-5, qDTFH-6A and qDTFH-6B), distributed across four LGs (1, 3, 5 and 6) (Fig. 4). Among these, one locus (qDTFH-6A) explained the majority of the variation (63 %), with the other five each contributing less than 10 % each (Table 2). This QTL showed a negative additive effect on days to flowering under high R/FR light quality, which indicated its effect was due to the allele from wild parent L. orientalis BGE 016880.

Fig. 4.

QTLs identified for flowering time under two different light quality environments (A, high R/FR; B, low R/FR) and flowering time sensitivity to this light quality change (C) in a lentil interspecific RIL population developed from a cross between L. culinaris cv. Lupa and L. orientalis BGE 016880. The horizontal dashed line on each graph represents the threshold LOD score for QTL identification after 1000 permutation tests.

Table 2.

Flowering DEGs from the contrast set between low red/far-red ratio and high red/far-red ratio for both L. orientalis BGE016880 and L. culinaris cv. Lupa and possible related QTLs identified from a lentil interspecific RIL population developed from L. culinaris cv. Lupa and L. orientalis BGE 016880. A gene was significantly differentially expressed if it had an adjusted P-value < 0.05 and |Log₂FC| ≥ 1.

Gene name	Chr no.	L. orientalis BGE 016880			L. culinaris cv. Lupa			Associated QTLs
		Adj.pval1	Log2FC2	Up.down	Adj.pval1	Log2FC2	Up.down
LcAGL6/13d	1	1.12E-09	3.05	Up-regulated	4.06E-05	1.33	Up-regulated	qDTFH-1A, qDTFL-1
LcSPL3a	2	4.86E-05	−2.72	Down-regulated	2.16E-03	−0.92	Non-DEGs
LcGA20oxC	2	1.48E-07	3.40	Up-regulated	2.56E-03	1.41	Up-regulated
LcAGL6/13c	2	2.20E-09	3.91	Up-regulated	1.05E-04	2.53	Up-regulated
LcGA3oxBC	2	2.10E-04	−1.95	Down-regulated	3.11E-02	−0.69	Non-DEGs
LcELF4a	3	6.27E-04	1.84	Up-regulated	5.55E-07	2.93	Up-regulated
PRR response regulator	3	3.74E-07	0.92	Non-DEGs	1.28E-10	1.53	Up-regulated
LcGA20oxG	3	3.97E-07	3.58	Up-regulated	4.57E-09	4.39	Up-regulated
LcTEMa	5	1.37E-02	1.73	Up-regulated	1.80E-03	1.83	Up-regulated
LcSVPb	5	1.42E-07	−1.65	Down-regulated	3.01E-10	−1.38	Down-regulated
LcFTb1/LcFTb2	6	3.76E-12	2.65	Up-regulated	2.48E-07	4.39	Up-regulated	qDTFH-6B, qDTFL-6B
LcFTa1	6	4.66E-03	1.62	Up-regulated	4.96E-04	0.23	Non-DEGs	qDTFH-6A, qDTFL-6A, qFTS-6
LcPIF6	6	6.97E-13	2.12	Up-regulated	2.04E-08	2.04	Up-regulated
LcSOC1b	7	3.89E-10	1.99	Up-regulated	1.28E-05	1.09	Up-regulated
LcTOC1a	7	4.86E-05	1.02	Up-regulated	8.62E-11	1.19	Up-regulated
LcFULb	7	1.24E-12	1.99	Up-regulated	1.60E-14	2.29	Up-regulated

¹Adj.pval: FDR-adjusted P-value. P-values of multiple testing were adjusted with the Benjamini–Hochberg (BH) procedure to correct false discovery rate (FDR) (Benjamini and Hochberg, 1995).

²Log₂FC: the log₂ fold change of gene abundance based on contrast groups.

Under low R/FR light conditions, we detected only four QTLs for flowering time, across LG1 (qDTFL-1) and LG6 (qDTFL-6A, qDTFL-6B, qDTFL-6C) (Fig. 4). As in high R/FR conditions, the overall effect was dominated by one locus (qDTFL-6A), which explained 46 % of the variance, with the three other loci contributing between 7 and 15 % each (Table 1). Again, the wild BGE 016880 allele contributed to the effect at this major locus.

Table 1.

QTLs controlling flowering time under different light quality environments and flower time sensitivity of a lentil interspecific recombinant inbred line population developed from L. culinaris cv. Lupa and L. orientalis BGE 016880

QTL name	Trait name	LG	Position (cM)	Left CI − Right CI (cM)1	Left marker1	Right marker1	LOD2	PVE2 (%)	Add2
qDTFH-1A	DTFH	1	154	153.5–154.5	Chr1_314467432	Chr1_314826609	12.8	7.3	2.95
qDTFH-1B	DTFH	1	675	673.5–677.5	Chr1_531417363	Chr1_531478466	4.0	1.7	1.46
qDTFH-3	DTFH	3	929	927.5–929.5	Chr3_408832764	Chr3_412274319	15.1	9.1	3.30
qDTFH-5	DTFH	5	607	606.5–607.5	Chr5_374745244	Chr5_360592403	6.8	3.2	1.99
qDTFH-6B	DTFH	6	19	18.5–20.5	Chr6_3105425	Chr6_2116708	14.0	8.1	3.10
qDTFH-6A	DTFH	6	398	397.5–398.5	Chr6_306050653	Chr6_306189556	42.0	62.9	−8.80
qDTFL-1	DTFL	1	188	187.5–188.5	Chr1_316264583	Chr1_332134411	9.5	7.6	1.74
qDTFL-6B	DTFL	6	29	27.5–29.5	Chr6_1026902	Chr6_904559	12.6	10.9	2.09
qDTFL-6C	DTFL	6	260	259.5–261.5	Chr6_68502582	Chr6_201166536	15.1	14.5	2.40
qDTFL-6A	DTFL	6	407	406.5–407.5	Chr6_301233901	Chr6_294069348	27.5	45.7	−4.29
qFTS-2	FTS	2	913	912.5–913.5	Chr2_311687210	Chr2_311337314	4.8	7.9	0.02
qFTS-3	FTS	3	888	886.5–888.5	Chr3_402268033	Chr3_404644546	4.3	7.2	0.02
qFTS-6	FTS	6	396	394.5–396.5	Chr6_310079803	Chr6_306019917	21.4	59.7	−0.05

DTFH, flowering time under high red/far-red light quality environment; DTFL, flowering time under low red/far-red light quality environment; FTS, flower time sensitivity towards different light quality environments.

¹Left, right CI refers to the left and right positions of confidence interval with one-LOD drop from the QTL; Left, right markers refers to the markers at the left, right CI position (cM).

²LOD: logarithm of odds; PVE: percentage of phenotypic variance explained by the QTL; Add: additive effect of the QTL. A positive or negative additive effect indicates the effect is contributed by an allele from L. culinaris cv. Lupa or L. orientalis BGE 016880, respectively.

Analysis of the flowering time response to R/FR (in terms of the FTS index) revealed three QTLs on LG2, 3 and 6 (qFTS-2, qFTS-3, qFTS-6) (Fig. 4). Locus qFTS-6 was located in a similar region of LG6 to qDTFH-6A and qDTFL-6A and explained 60 % of the phenotypic variance of the contribution from the wild BGE 016880 allele. The other two loci each explained < 10 % each.

The major locus for each of the three traits (i.e. DTFH, DTFL and FTS) overlapped on LG6 (Fig. 4), raising the possibility that these three loci might reflect the contribution of a single underlying gene. Across the three traits, three other pairs of loci showed similar close co-location: on LG1 (qDTFH1A and qDTFL1), LG3 (qDTFH-3 and qFTS-3) and LG6 (qDTFH-6B and qDTFL-6B), also suggesting a possible common genetic basis for the loci in each pair. Finally, four of the detected loci (one for each trait) were in unique positions on LG1 (qDTFH-1B), LG2 (qFTS-2), LG5 (qDTFH-5) and LG6 (qDTFL-6C). Overall, our analysis across both light quality conditions revealed a minimum of seven distinct QTLs contributing to the variability in DTF under these experimental conditions.

De novo assembly of transcriptome

From what we have done so far based on the genome assemblies of different Lens species and intensive genotyping of intra- and inter-specific lentil RIL populations, we know that a genome size difference exists among Lens species and chromosomal rearrangements are common even within cultivated lentil. BGE 016880 is an accession of L. orientalis while Lupa is a Spanish cultivar (Fratini et al., 2007). To remove bias that could result from using the L. culinaris cv. CDC Redberry reference genome for differential gene expression analysis, we carried out de novo assembly of the transcriptomes from RNAseq data for both BGE 016880 and Lupa.

A total of 705.1 Mb paired-reads of the transcriptomic data were generated using the Illumina HiSeq 2500 platform (paired end of 125 bp × 2) of which 359.4 Mb belonged to L. orientalis BGE 016880 and 345.7 Mb belonged to L. culinaris cv. Lupa (Supplementary File S4). After quality filtering and pre-processing of reads, 97.8 % of the reads for both L. orientalis BGE 016880 and L. culinaris cv. Lupa were kept for further analysis. These high-quality reads of 30 samples from five developmental stages were pooled together for L. orientalis BGE 016880 and L. culinaris cv. Lupa, respectively, and used for de novo transcriptome assembly. Trinity assembler generated 138 108 transcripts (78 707 raw genes) with an N50 of 2255 bp for L. orientalis BGE 016880, and 135 723 transcripts (76 595 raw genes) with an N50 of 2252 bp for L. culinaris cv. Lupa. (Supplementary Data File S1 – Table S3).

Differential gene expression analysis

De novo assembled transcriptomes of BGE 016880 and Lupa were used to quantify the abundance of the transcripts for all samples from BGE 016880 and Lupa respectively using the Salmon package (Patro et al., 2017). 3D RNA-seq pipeline (Calixto et al., 2018; Guo et al., 2019) was then used for the analysis of differential gene expression, and a detailed method can be found in Supplementary Data File S1 – Methods.

A total of 2573 and 1367 DEGs were obtained from the contrast sets between the low R/FR light environment and high R/FR environment for five different growth stages of L. orientalis BGE 016880 and L. culinaris cv. Lupa, respectively (Supplementary Data File S5). Of the 2573 DEGs from BGE 016880, 933, 1061, 64, 139 and 376 DEGs were observed in the T1, T2, T3, T4 and T5 stage, respectively (Supplementary Data File S1 – Fig. S3A). Of the 1367 DEGs from Lupa, 434, 84, 232, 362 and 255 DEGs were observed in these five stages, respectively (Supplementary Data File S1 – Fig. S3B). Up- and down-regulated genes at these five stages were quite different for both L. orientalis BGE 016880 and L. culinaris cv. Lupa (Fig. 5), but most DEGs were identified about 2 weeks before flowering, which corresponds to the transition from vegetative growth to reproductive growth for both BGE 016880 and Lupa. When samples from the five development stages were pooled together to look at the DEGs between low R/FR and high R/FR light quality, 297 and 156 were identified from L. orientalis BGE 016880 and L. culinaris cv. Lupa, respectively (Supplementary Data File S5). DEGs in L. orientalis BGE 016880 were almost double those in L. culinaris cv. Lupa between low R/FR and high R/FR light quality environments.

Fig. 5.

Alluvial plots showing up-regulated genes (red flow), non-differentially expressed genes (green flow) and down-regulated genes (blue flow) from the contrast between low R/FR and high R/FR from five different growth stages for L. orientalis BGE016880 (A) and L. culinaris cv. Lupa (B).

Identification of flowering DEGs

We further examined the DEGs for possible functional links to flowering time regulation. To do this, we used the list of 244 lentil genes that we curated based on the reference genome (Supplementary File S2) and examined their representation within the DEG set using a custom Perl script (Supplementary File S6). This analysis identified a total of 82 genes as candidate flowering-related genes among the DEG sets identified for L. orientalis BGE 016880 and 62 from the L. culinaris cv. Lupa DEG sets. The number of flowering-related DEGs associated with each of the five development stages were as follows: T1 – 15, T2 – 24, T3 – 13, T4 – 12 and T5 – 18 for BGE 016880. When sorting flowering-related DEGs within the development stages for Lupa, we saw the following numbers: T1 – 10, T2 – 4, T3 – 15, T4 – 17 and T5 – 16. Flowering-related DEGs again showed stage-specific regulation in both BGE 016880 and Lupa.

Since the flowering times for both BGE 016880 and Lupa were significantly different and spread from 27 to 60 d after emergence (Supplementary Data File S1 – Fig. S1), we decided to look at flower DEGs from the DEG set of samples from five development stages, instead of any single stage, to understand the flowering time regulation between low R/FR and high R/FR light quality environments. Seventeen flower DEGs were identified in both BGE 016880 and Lupa, out of 297 and 156 DEGs respectively. Of these, 12 were common to both BGE 016880 and Lupa, while five others were unique to the DEG set for one or the other line (Table 2).

Verification of flowering DEGs through RT-qPCR

We next validated results from RNAseq analysis by RT-qPCR, for selected genes including flowering DEGs in the QTL regions, non-DEG flowering genes within the QTL regions, and flowering DEGs that were not within QTL regions. From the subsets of flowering-related DEGs, we selected five genes that were located within or close to the QTL confidence intervals (Table 2). The major loci qDTFH-6A, qDTFL-6A and qFTS-6 in the central region of LG6 co-located with three FT homologues: LcFTa1, LcFTa2 and LcFTc (Lcu.2RBY.6g043850, Lcu.2RBY.6g043870, Lcu.2RBY.6g043940), and a light-regulated transducin/WD-like repeat-protein (LWD) gene orthologue LcLWD1 (Lcu.2RBY.6g043520). Of these, LcFTa1 was the only one represented in the DEG set in the wild parent, but not in the cultivated parent. Another cluster of FT homologues LcFTb1 and LcFTb2 (Lcu.2RBY.6g000730 and Lcu.2RBY.6g000760) were located near the qDTFH-6B and qDTFL-6B loci at the top of LG6. A MADS-box gene in the AGAMOUS-LIKE 6 (AGL6) clade (Lcu.2RBY.1g040350/LcAGL6/13d) was the closest flowering-related DEG to the loci qDTFH-1A and qDTFL-1. We also selected one gene, LcELF4a (Lcu.2RBY.3g037650), a flowering-related DEG not located within or near any QTL region.

The results were consistent with and further supported our RNAseq results (Fig. 6). For the major QTL region in the middle of LG6, we confirmed the strong upregulation of LcFTa1 in response to low R/FR in L. orientalis BGE 016880 but not in L. culinaris cv. Lupa. In the same region, LcLWD1 and LcFTa2 had lower (or no) expression in L. orientalis BGE 016880, but similar expression level in L. culinaris cv. Lupa under different light quality environments, whereas LcFTc expression was unaffected by R/FR (Fig. 6). For the QTL region at the top of LG6, the DEGs FTb1 and FTb2 showed consistent increased expression under low R/FR light quality in both L. orientalis BGE 016880 and L. culinaris cv. Lupa. LcELF4a was not within or near any QTL region, but it was identified as a flowering DEG, and RT-qPCR results also supported this. This could mean that additional loci with small effects were not identified, a common limitation of QTL mapping (Myles and Wayne, 2008).

Fig. 6.

Relative expression of flowering-related genes from the samples grown at low red/far-red ratio and high red/far-red ratio from five different growth stages for both L. orientalis BGE016880 and L. culinaris cv. Lupa assessed using RT-qPCR. Sample collection started 2 weeks after emergence and continued once a week for 5 weeks. Samples were taken at the same time of day for each collection. Actin was used as the reference gene for normalization of the data and the delta-delta CT method was used to derive the relative gene expression value 2^-ΔΔCT using samples from the T1 stage at the low R/FR light quality environment as reference. Values represent the mean of biological replicates with their corresponding standard deviation.

DISCUSSION

Optimization of seed yield and reproductive success in nature and in agriculture depends to a large extent on the appropriate seasonal control of flowering (Jung and Müller, 2009). Light is an important daily, seasonal and spatial variable for plants, and flowering time is not only affected by light duration (photoperiod), but also by light quality (Adams et al., 2009; Casal, 2013). In particular, a reduction in R/FR occurs in light reflected from green plant tissue, provides cues about prospective competition from neighbouring plants and leads to a set of morphological responses including an acceleration of flowering (Whitelam and Devlin, 1997). We have previously reported species and genotype differences in response to light quality changes within the genus Lens, where some wild lentils seemed to be less sensitive than cultivated types and therefore may perform better under variable light quality (Yuan et al., 2017).

The parents of the interspecific RIL population used in this study have contrasting flowering responses to changes in light quality, with the wild parent, BGE 016880, being less sensitive to changes than the cultivated parent, Lupa. As expected, the RILs within the population exhibited segregation for flowering time under different light quality environments, as well as flowering time sensitivity towards different light qualities. The presence of less sensitive RILs than the wild parent BGE 016880, and more sensitive RILs than the cultivated parent, suggests that alleles from both parents have contributed to the response, and rearrangement of alleles led to transgressive segregation in the RIL population.

In our current study, 4073 SNP markers were distributed across six LGs with an average marker density of 1.5 cM. The physical size of the lentil genome is estimated to be 4086 Mb (Arumuganathan and Earle, 1991). Based on the distance of the map developed in this study, an average of 0.69 Mb cM^–1 was covered. Some similar linkage maps generated in lentil RIL populations had average coverages of 5.9 Mb cM^–1 (Fedoruk et al., 2013), 2.13–3.14 Mb cM^–1 (Sudheesh et al., 2016) or 1.01 Mb cM^–1 (Ates et al., 2016) using SNP markers generated from other approaches. It appears that the number of RADseq-generated SNP markers used in the current map may help to improve coverage of the lentil genome.

There are few reports on flowering time-related QTL studies in lentil, and most have used different marker systems, making it difficult to find consensus across studies (Sarker et al., 1999; Fratini et al., 2007; Tullu et al., 2008; Fedoruk et al., 2013). So far, there have been no reported studies on the genetics behind flowering time and light quality change in lentil, although the identity of the major Sn locus as an orthologue of the circadian clock gene ELF3a (Sarker et al., 1999; Weller et al., 2012) suggests a possible influence on both photoperiod and light quality sensitivity (Jiménez-Gómez et al., 2010; Weller et al., 2012). In our current study, 13 QTLs were identified in the RIL population for days to flowering under high R/FR light quality, days to flowering under low R/FR light quality and flowering time sensitivity towards light quality change.

Studies using comparative genetics as well as functional genomics have shown that most major flowering time genes are well conserved between Arabidopsis and a large range of crop species including legumes (Hecht et al., 2005; Roux et al., 2006; Kim et al., 2013; Weller and Ortega, 2015). Using 244 in-house curated lentil homologues of Arabidopsis thaliana and Medicago truncatula flowering-related genes, we were able to look at the corresponding QTL regions in the lentil reference genome and associated them with flower DEGs. Three major QTLs identified in the study were within a similar region in the middle of LG6, where a tandem array of florigen (FT) genes, LcFTa1, LcFTa2 and LcFTc, as well as a light-regulated transducin/WD-like repeat-protein (LcLWD1), are located, although only one (LcFTa1) showed significant differential expression within the region. These three QTLs explained more than 45 % of the phenotypic variance and were in a region syntenic with a section of Medicago chromosome 7 and pea linkage group V, containing a similar tandem array of FTa1, FTa2 and FTc genes (Hecht et al., 2005; Weller and Ortega, 2015). This region is related to control of flowering time in several temperate legumes, including pea (Hecht et al., 2011), M. truncatula (Laurie et al., 2011), faba bean (Cruz-Izquierdo et al., 2012), narrow-leafed lupin (Nelson et al., 2017) and chickpea (Ortega et al., 2019). When de novo assembled transcriptomes of L. orientalis BGE016880 and L. culinaris cv. Lupa were aligned with our reference genome, the LcFTa2 transcript could not be detected in L. orientalis BGE016880, implying a possible deletion, and suggesting that this gene is probably not involved in the response to changes in light quality in this interspecific RIL population. Interestingly, a similar deletion was observed to associate with a QTL conferring early flowering in chickpea (Ortega et al., 2019), but was also not considered likely to explain the effect, in view of the generally promotive effects of FT genes on flowering (Laurie et al., 2011; Hecht et al., 2011).

Another cluster of QTLs in a distinct region at the top of LG6 was shown to co-locate with two other differentially expressed flowering genes, the florigen genes LcFTb1 and LcFTb2. In both pea and Medicago, orthologues of these genes have also been implicated in flowering time control through genetic and gene expression studies (Hecht et al., 2011; Laurie et al., 2011; Ridge et al., 2017; Putterill et al., 2019; Jaudal et al., 2020). As a flowering pathway integrator, Flowering Locus T genes promote the transition to flowering and play a key role in plant adaptation and crop improvement (Wickland and Hanzawa, 2015). In the related temperate legume subterranean clover, FTb2 has also been implicated in the acceleration of flowering by low R/FR through gene expression studies (Pazos-Navarro et al., 2018).

Two other QTLs identified were within the same region on LG1 where Agamous-like MADS-box protein AGL6/13d (LcAGL6/13d) was the sole flowering-related DEG. AGL6 represents a subfamily of MADS-box transcription factor genes that have various roles regarding flowering time and flower development (Dreni and Zhang, 2016). In Arabidopsis, the activation of AGL6 is associated with the down-regulation of floral repressor FLC/MAF genes, and up-regulation of the floral promoter FT (Yoo et al., 2011). RT-qPCR analysis confirmed that LcAGL6/13d consistently showed high expression under low R/FR light quality compared to high R/FR light quality in both L. orientalis BGE016880 and L. culinaris cv. Lupa.

FT regulators such as Phytochrome Interacting Factors (PIFs) and Short Vegetative Phase (SVP) target FT promoters and non-coding regions. These regulators were also differentially expressed in our study. PIFs belong to a class of basic helix–loop–helix (bHLH) transcription factors and are involved in promoting the transition to flowering by acting upstream of FT and TWIN SISTER of FT (TSF) under low R/FR in Arabidopsis (Galvāo et al., 2019). In soybean, PIF3 has been identified as a shade-responsive gene through an RNAseq experiment (Horvath et al., 2015). SVP, or Agamous-like MADS-box protein AGL22, is a negative regulator of flowering in Arabidopsis (Hartmann et al., 2000) and can bind to the promoters of SOC1 and FT for transcriptional repression (Li et al., 2008). SVP expression decreased after gibberellin (GA) treatment, which suggested the involvement of GA on flowering could be partly mediated through SVP (Li et al., 2008). Two other members of MADS-box transcription factor genes, Suppressor of overexpression of constans 1 (SOC1) and FRUITFULL (FUL), were also identified as flowering-related DEGs in our study. As a floral pathway integrator, SOC1 (Agamous-like MADS-box protein AGL20) incorporates multiple flowering signals to promote flowering and can be induced by GA, while repressed by FLC and SVP in Arabidopsis (Lee and Lee, 2010). SOC1 works downstream of FT to promote flowering through up-regulation of the gene LEAFY, which links floral induction and floral development. A recent study in M. truncatula showed that MtSOC1 genes can be induced by the MtFTa1 gene and function redundantly to accelerate flowering (Fudge et al., 2018). FUL, or Agamous-like MADS-box protein AGL8, has been shown to work together with SOC1, but act against the effects of FLC and SVP to promote flowering (Balanzà et al., 2014). The subset of flowering-related genes differentially expressed under low compared to high R/FR light quality was very similar in both L. orientalis BGE016880 and L. culinaris cv. Lupa, indicating a high similarity in the underlying regulatory processes. However, they differed to some extent both in the level of the expression differences and the stage at which specific differences were detected.

Overall, our results from QTL analysis and gene expression point most clearly to FTa1 as a probable basis for the observed differences in flowering sensitivity to light quality between L. orientalis BGE016880 and L. culinaris cv. Lupa. Similar parallel evidence suggests a weaker role for FTb1/2 genes, and identifies an AGL6/13-like MADS-box transcription factor as a potential candidate for a third QTL. In addition, the wider gene expression dataset suggests that a number of other genes including MADS box genes LcSVPb, LcSOC1b and LcFULb, the bHLH transcription factor LcPIF6, and genes related to gibberellin synthesis LcGA20oxC,G, may be involved in the light quality response in leaf tissue, either upstream or downstream of FT genes (Fig. 7). An important characteristic of the flowering regulatory network is that all signalling pathways respond to different endogenous and environmental signals but work together through a delicate and complicated balance to adjust the switch to flower in a changing environment. Our results suggested that FT genes together with various MADS-box and bHLH transcription factors played an important role in flowering time regulation under different light quality environments. The identified QTLs and candidate genes provide a foundation for breeding for better adaptation to variable light environments, which could lead to cultivars with improved competitiveness under weed pressure or in intercropping situations.

Fig. 7.

A model for the role and interactions of lentil flower genes under low R/FR light quality environment. This model summarizes the major results from this study, and the hypothetical interactions are based on previous network studies on Arabidopsis and other legumes (Putterill et al., 2013; Blümel et al., 2015; Weller and Ortega, 2015; Chen et al., 2018). Arrows indicate a promoting interaction, a T-end indicates an inhibiting interaction and a straight line marks an interaction with no firm direction. A gene name shown in blue (LcFTa1) was a DEG only in L. orientalis BGE 016880.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. File S1: Figures, tables and methods. File S2: Lentil flowering homologues. File S3: Map matrix file. File S4: RNAseq data. File S5: DEG list. File S6: Trinity DEG list.

FUNDING

The research was conducted as part of the AGILE project (Application of Genomics to Innovation in the Lentil Economy), a Genome Prairie managed project funded by Genome Canada, Western Grains Research Foundation, the Province of Saskatchewan and the Saskatchewan Pulse Growers.