2. MATERIALS AND METHODS

2.1 Plant materials and light regimens

The study was performed in two sequential growth cycles as complete replicates. In the first growth cycle, we mainly counted the number of floret primordia with different morphologies and the number of grains (Figure S1); and in the second growth cycle, we monitored the growth, development, and transcriptional dynamics of all floret primordia in the spike. Spring wheat seeds (cv. longfumai1) were selected, washed, stratified, and germinated. The germination seeds with similar radicle length were sown in pots (volume 14.7 L, 30 seeds per pot) filled with a 2: 1: 1 (v/v/v) mixture of garden soil, humus, and sand. The pots were transferred to a fully enclosed plant factory (a type of controlled environment agriculture), that was programmed to run at a temperature of 20-21oC during 06:00-18:00 and 15-16 oC during 18:00-next day 06:00, with a relative humidity of 60-70% and a CO2 level of 380-400 ppm. There were 30 pots, and plants were watered every two days with the modified Hoagland’s nutrient solution (Table S1).
The white light-emitting diodes (LEDs) were used to provide 12 h of illumination per day (06:00-18:00) and the photosynthetic photon flux density (PPFD) was adjusted to ~320 µmol m−2 s−1 at the top of the plant canopy. When plants reached the terminal spikelet stage, the pots were divided into three groups for different light treatments. Light regimens included: (i) SW (12 h short photoperiod supplemented with white LEDs, light (06:00-18:00)/ dark (18:00-next day 06:00)); (ii) LW (22 h long photoperiod supplemented with white LEDs, light (06:00-next day 04:00)/ dark (04:00-06:00)); (iii) LR (22 h long photoperiod supplemented with red LEDs, light (06:00-next day 04:00)/ dark (04:00-06:00)). The photoperiod extension was performed at low intensity (approximately 12 µmol m−2 s−1 PPFD). Photoperiod and light spectral quality were measured with a portable spectroradiometer (Plant Lighting Analyzer, v. 2.00, China) and were showed in Figure 1 and Table S2.

2.2 Floret primordia number and morphology measurements

Representative six main spikes per treatment were collected every three days to count the number of total, fertile, and infertile floret primordia per spike and per spikelet according to the different morphologies (Figure S1b). Photomicrographs of each floret primordium within a particular central spikelet were taken with a microscope (MZ101, Mshot, China) equipped with a digital camera (MSX1, Mshot, China).

2.3 Dry matter accumulation and sugar measurement

Representative six plants per treatment were collected every three days and dried (at 105 oC for 10 min and then at 70oC for 48 h) to a constant weight to determine the dry matter accumulation and partitioning of plants and spikes. Dry spike samples were used for soluble sugar extraction (Hanft & Jones, 1986). Soluble sugar content was determined following the anthrone-sulfuric acid method (Yemm & Willis, 1954). Glucose and fructose content were determined by enzymatic analysis as described by Blunden & Wilson (1985).

2.4 Samples preparation and RNA sequencing

Pooled samples from at least three spikes every three or six days were collected within two to five hours after the illuminate initiation, then immediately frozen in liquid nitrogen and stored at -80oC until further used for RNA extraction. Three biological replicates were used for RNA sequencing.
Total RNA was isolated and purified using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturer’s protocol. The RNA purity and integrity were quantified and assessed using the NanoDrop ND-1000 (NanoDrop, Wilmington, DE, USA) and the Bioanalyzer 2100 (Agilent, CA, USA), respectively. RNA-seq libraries were constructed according to the manufacturer’s protocol of the NEBNext mRNA-seq library preparation kit, and paired-end sequencing libraries with an insert size of ~300 bp were sequenced on an Illumina Novaseq™ 6000 platform (LC-Bio Technology CO., Ltd., Hangzhou, China) following the vendor’s recommended protocol. Clean reads were mapped to the International Wheat Genome Sequencing Consortium (IWGSC) RefSeq version 2.1 genome assembly using Hisat2-2.0.4 software (Kim et al. , 2015). The expression level of each gene was normalized to fragments per kilobase of exon model per million mapped reads (FPKM). Comparisons of biological replicates showed that their expression levels were highly correlated (R > 0.95; Figure S2). Hence, a gene was considered expressed if it had an average FPKM ≥ 1. Differential expression analyses between treatments were performed using the platform OmicStudio (https://www.omicstudio.cn) according to edgeR, and the differentially expressed genes (DEGs) were selected with an absolute log2 fold change (FC) ≥ 1, a statistical significance (false discovery rate (FDR)-adjusted P < 0.05), and FPKM ≥ 2 in highly expressed groups. The principal component analysis (PCA), hierarchical clustering, and Venn diagram were performed using the anosim and prcomp function, the hclust function, and the vennDiagram function, respectively, in R software. The Heat maps were drawn using the TBtools (Chen et al. , 2020).

2.5 Gene coexpression and functional enrichment analysis

The k -means method in R software was used for coexpression analysis of DEGs selected in LR versus LW. Functional category of each coexpression module was performed using Gene Ontology (GO) analysis in the platform OmicStudio (https://www.omicstudio.cn). Significantly enriched GO terms met the condition of FDR < 0.05.

2.6 Identification of candidate key genes

A web database, STRING (https://cn.string-db.org/) (Szklarczyk et al. , 2019), was used for the protein-protein interaction (PPI) analysis, specifying Triticum aestivum as the target organism with a minimum required interaction score of 0.4. The PPI network was then sent to Cytoscape (v. 3.8.2) (Shannon et al. , 2003), and the Cytohubba module and Degree method were used to screen out the five hub nodes. Genes encoding proteins with higher degree centrality scores were regarded as more important hubs in the network.

2.7 Statistical analysis

All experimental data was processed using Microsoft Excel 2010 (Microsoft Corporation, USA). One-way analysis of variance (ANOVA) and LSD’s post hoc test (P < 0.05) were performed for statistical analysis of the phenotypes. Figures were created using OriginPro 2017 (OriginLab Corp., Northampton, MA, USA), R (v. 4.1.2), and Adobe Illustrator CC 2019 (Adobe Inc., USA).